diff --git a/.agents/.gitignore b/.agents/.gitignore
index 93c0f73..2cf3fac 100644
--- a/.agents/.gitignore
+++ b/.agents/.gitignore
@@ -1 +1,2 @@
 settings.local.json
+.nicc*
diff --git a/.github/workflows/README.md b/.github/workflows/README.md
index 5e21570..dd77756 100644
--- a/.github/workflows/README.md
+++ b/.github/workflows/README.md
@@ -10,7 +10,7 @@ Runs on every push and pull request to main branches. Includes:
 
 - **unit-tests**: Cross-platform unit tests
   - Runs on Ubuntu and Windows
-  - Python 3.10, 3.11, and 3.12
+  - Python 3.11 and 3.12
   - Uses PyTorch CPU version to avoid GPU dependencies
   - Excludes slow tests and tests requiring external data
   - Generates coverage reports
@@ -83,7 +83,7 @@ The workflows use multiple caching layers to speed up builds:
 
 ### Important: GPU Tests Are Disabled by Default
 
-⚠️ **GPU tests do NOT run automatically** to prevent jobs from waiting indefinitely in the queue when no runner is available.
+**GPU tests do NOT run automatically** to prevent jobs from waiting indefinitely in the queue when no runner is available.
 
 To run GPU tests, you must either:
 1. **Manually trigger the workflow**: Go to Actions > CI > Run workflow
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 5ded2fa..4e656ff 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -30,7 +30,7 @@ jobs:
       fail-fast: false
       matrix:
         os: [ubuntu-latest, windows-latest]
-        python-version: ['3.10', '3.11', '3.12']
+        python-version: ['3.11', '3.12']
 
     steps:
     - name: Checkout code
@@ -79,7 +79,9 @@ jobs:
           libxrender1 \
           libxext6 \
           libxrandr2 \
-          libxi6
+          libxi6 \
+          xvfb \
+          libosmesa6
         sudo apt-get clean
         sudo rm -rf /var/lib/apt/lists/*
 
@@ -99,13 +101,20 @@ jobs:
       run: |
         pip list
 
-    - name: Run unit tests (fast, no external data)
+    - name: Run unit tests (fast, no external data) - Ubuntu
+      if: matrix.os == 'ubuntu-latest'
+      run: |
+        xvfb-run -a --server-args="-screen 0 1024x768x24" \
+          pytest tests/ -v -m "not slow and not requires_data and not experiment" --cov=physiomotion4d --cov-report=xml --cov-report=term --cov-report=html
+
+    - name: Run unit tests (fast, no external data) - Windows
+      if: matrix.os == 'windows-latest'
       run: |
         pytest tests/ -v -m "not slow and not requires_data and not experiment" --cov=physiomotion4d --cov-report=xml --cov-report=term --cov-report=html
 
     - name: Upload coverage to Codecov
       uses: codecov/codecov-action@v4
-      if: matrix.os == 'ubuntu-latest' && matrix.python-version == '3.10'
+      if: matrix.os == 'ubuntu-latest' && matrix.python-version == '3.11'
       with:
         token: ${{ secrets.CODECOV_TOKEN }}
         file: ./coverage.xml
@@ -115,7 +124,7 @@ jobs:
 
     - name: Upload coverage artifacts
       uses: actions/upload-artifact@v4
-      if: matrix.os == 'ubuntu-latest' && matrix.python-version == '3.10'
+      if: matrix.os == 'ubuntu-latest' && matrix.python-version == '3.11'
       with:
         name: coverage-report-unit-tests
         path: htmlcov/
@@ -148,10 +157,10 @@ jobs:
         echo "Disk space after cleanup:"
         df -h
 
-    - name: Set up Python 3.10
+    - name: Set up Python 3.11
       uses: actions/setup-python@v5
       with:
-        python-version: '3.10'
+        python-version: '3.11'
         cache: 'pip'
         cache-dependency-path: pyproject.toml
 
@@ -185,7 +194,9 @@ jobs:
           libxrender1 \
           libxext6 \
           libxrandr2 \
-          libxi6
+          libxi6 \
+          xvfb \
+          libosmesa6
         sudo apt-get clean
         sudo rm -rf /var/lib/apt/lists/*
 
@@ -218,17 +229,20 @@ jobs:
 
     - name: Run USD conversion tests
       run: |
-        pytest tests/test_convert_vtk_to_usd_polymesh.py -v -m "not slow" --cov=physiomotion4d --cov-append --cov-report=xml
+        xvfb-run -a --server-args="-screen 0 1024x768x24" \
+          pytest tests/test_convert_vtk_to_usd_polymesh.py -v -m "not slow" --cov=physiomotion4d --cov-append --cov-report=xml
       continue-on-error: true
 
     - name: Run USD utility tests
       run: |
-        pytest tests/test_usd_merge.py tests/test_usd_time_preservation.py -v --cov=physiomotion4d --cov-append --cov-report=xml
+        xvfb-run -a --server-args="-screen 0 1024x768x24" \
+          pytest tests/test_usd_merge.py tests/test_usd_time_preservation.py -v --cov=physiomotion4d --cov-append --cov-report=xml
       continue-on-error: true
 
     - name: Run all integration tests
       run: |
-        pytest tests/ -v -m "not slow and not experiment and not requires_gpu"
+        xvfb-run -a --server-args="-screen 0 1024x768x24" \
+          pytest tests/ -v -m "not slow and not experiment and not requires_gpu"
       continue-on-error: true
 
     - name: Upload coverage to Codecov
@@ -345,10 +359,10 @@ jobs:
       with:
         lfs: true
 
-    - name: Set up Python 3.10
+    - name: Set up Python 3.11
       uses: actions/setup-python@v5
       with:
-        python-version: "3.10"
+        python-version: "3.11"
         cache: 'pip'
 
     - name: Install dev dependencies
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index c750323..2ea46a0 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -32,7 +32,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
         with:
-          python-version: '3.10'
+          python-version: '3.11'
 
       - name: Install build tools
         run: |
diff --git a/CLAUDE.md b/CLAUDE.md
index b045f66..7ec2ae5 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -36,7 +36,7 @@ py -m pytest tests/ --ignore=tests/test_segment_chest_total_segmentator.py \
 # With coverage
 py -m pytest tests/ --cov=src/physiomotion4d --cov-report=html
 
-# Experiment notebook tests (very slow, opt-in)
+# Experiment script tests (very slow, opt-in)
 py -m pytest tests/ --run-experiments
 
 # Create missing baselines
diff --git a/README.md b/README.md
index 8a605e3..b33d13b 100644
--- a/README.md
+++ b/README.md
@@ -28,7 +28,7 @@ includes:
 > research and visualization toolkit. It is not a medical device and must not
 > be used for diagnosis, treatment planning, or clinical decision-making.
 
-## 🚀 Key Features
+## Key Features
 
 - **Complete 4D Medical Imaging Pipeline**: End-to-end processing from 4D CT data to animated USD models
 - **Multiple AI Segmentation Methods**: TotalSegmentator and Simpleware cardiac segmentation
@@ -37,7 +37,7 @@ includes:
 - **Physiological Motion Analysis**: Capture and visualize cardiac and respiratory motion
 - **Flexible Workflow Control**: Step-based processing with checkpoint management
 
-## 📋 Supported Applications
+## Supported Applications
 
 - **Cardiac Imaging**: Heart-gated CT processing with cardiac motion analysis
 - **Pulmonary Imaging**: Lung 4D-CT processing with respiratory motion tracking
@@ -45,11 +45,11 @@ includes:
 - **Research Visualization**: Advanced medical imaging research in Omniverse
 - **Clinical Planning**: Dynamic anatomical models for treatment planning
 
-## 🛠️ Installation
+## Installation
 
 ### Prerequisites
 
-- Python 3.10+ (Python 3.10, 3.11, or 3.12 recommended)
+- Python 3.11+ (Python 3.11 or 3.12 recommended)
 - NVIDIA GPU with CUDA 13 — recommended for production use; CPU-only installation is supported but slow
 - 16GB+ RAM (32GB+ recommended for large datasets)
 - NVIDIA Omniverse (for USD visualization)
@@ -109,7 +109,7 @@ print(f"PhysioMotion4D version: {physiomotion4d.__version__}")
 print(WorkflowConvertHeartGatedCTToUSD.__name__)
 ```
 
-## 🏗️ Package Architecture
+## Package Architecture
 
 ### Core Components
 
@@ -156,7 +156,7 @@ print(WorkflowConvertHeartGatedCTToUSD.__name__)
 
 ## Getting Started: Tutorials
 
-The `tutorials/` directory contains six end-to-end Python scripts, one for each
+The `tutorials/` directory contains nine end-to-end Python scripts, one for each
 major workflow. They are the recommended starting point for new users.
 
 | # | Script | Workflow | Dataset |
@@ -167,17 +167,21 @@ major workflow. They are the recommended starting point for new users.
 | 4 | `tutorials/tutorial_04_fit_statistical_model_to_patient.py` | Fit statistical model to patient | KCL-Heart-Model plus Tutorial 3 output |
 | 5 | `tutorials/tutorial_05_vtk_to_usd.py` | VTK surfaces to animated USD | output of tutorial 2 |
 | 6 | `tutorials/tutorial_06_reconstruct_highres_4d_ct.py` | Reconstruct high-res 4D CT | DirLab-4DCT (manual) |
+| 7 | `tutorials/tutorial_07_dirlab_pca_model.py` | Build a surface PCA lung-lobe model and fit all cases | DirLab-4DCT (manual) |
+| 8 | `tutorials/tutorial_08_dirlab_pca_time_series.py` | Propagate PCA-fitted lung-lobe meshes through DirLab time series | DirLab-4DCT plus Tutorial 7 output |
+| 9 | `tutorials/tutorial_09_physicsnemo_mesh_stage_model.py` | Train a PhysicsNeMo mesh stage model | Tutorial 8 output |
 
-Each script is runnable directly:
+Each tutorial is a `# %%` percent-cell Python script. Paths are defined near
+the top of the script; edit those constants for custom data/output locations,
+or use the installed `physiomotion4d-*` CLI commands when you want path
+arguments.
 
 ```bash
 # Tutorial 1 (CPU-safe ANTs registration; requires Slicer-Heart-CT data)
-python tutorials/tutorial_01_heart_gated_ct_to_usd.py \
-    --data-dir ./data --output-dir ./output/tutorial_01
+python tutorials/tutorial_01_heart_gated_ct_to_usd.py
 
 # Tutorial 2 (CT to VTK)
-python tutorials/tutorial_02_ct_to_vtk.py \
-    --data-dir ./data --output-dir ./output/tutorial_02
+python tutorials/tutorial_02_ct_to_vtk.py
 ```
 
 See `tutorials/README.md` for the full tutorial index, dataset preparation
@@ -189,7 +193,7 @@ This quickstart uses the public Slicer-Heart 4D CT sample. Data downloading and
 a CUDA-capable GPU are required for practical runtime.
 
 ```bash
-python -c "import pathlib, urllib.request; pathlib.Path('data/test').mkdir(parents=True, exist_ok=True); urllib.request.urlretrieve('https://github.com/SlicerHeart/SlicerHeart/releases/download/TestingData/TruncalValve_4DCT.seq.nrrd', 'data/test/TruncalValve_4DCT.seq.nrrd')"
+python -c "from physiomotion4d import DataDownloadTools; DataDownloadTools.DownloadSlicerHeartCTData('data/test')"
 
 physiomotion4d-heart-gated-ct data/test/TruncalValve_4DCT.seq.nrrd \
     --registration-method ants \
@@ -197,7 +201,7 @@ physiomotion4d-heart-gated-ct data/test/TruncalValve_4DCT.seq.nrrd \
     --project-name slicer_heart_quickstart
 ```
 
-## 🎯 Quick Start
+## Quick Start
 
 ### Command-Line Interface
 
@@ -326,11 +330,14 @@ segmenter = SegmentChestTotalSegmentator()
 image = itk.imread("chest_ct.nrrd")
 masks = segmenter.segment(image, contrast_enhanced_study=True)
 
-# Extract individual anatomy masks by key
-heart_mask = masks["heart"]
-vessels_mask = masks["major_vessels"]
-lungs_mask = masks["lung"]
+# Result always contains "labelmap" plus one entry per anatomy group the
+# segmenter registered (heart, lung, bone, major_vessels, soft_tissue,
+# contrast, other for SegmentChestTotalSegmentator). The exact key set is
+# segmenter-specific; check membership when targeting multiple segmenters.
 labelmap = masks["labelmap"]
+heart_mask = masks["heart"]
+if "lung" in masks:
+    lungs_mask = masks["lung"]
 ```
 
 ### Image Registration
@@ -371,17 +378,22 @@ PhysioMotion4D provides two APIs for converting VTK data to USD for NVIDIA Omniv
 #### Option 1: High-Level ConvertVTKToUSD (for PyVista/VTK objects)
 
 ```python
-from physiomotion4d import ConvertVTKToUSD
+from physiomotion4d import ConvertVTKToUSD, SegmentChestTotalSegmentator
 import pyvista as pv
 
 # Load VTK data
 meshes = [pv.read(f"cardiac_frame_{i:03d}.vtp") for i in range(20)]
 
-# Convert to animated USD with anatomical labels
+# Convert to animated USD with anatomical labels. Pass `segmenter` so the
+# converter groups labeled prims by anatomy type:
+#   /World/CardiacModel/heart/<organ>, /World/CardiacModel/lung/<organ>, ...
+# Without `segmenter`, all labeled prims land under /World/CardiacModel/Anatomy.
+seg = SegmentChestTotalSegmentator()
 converter = ConvertVTKToUSD(
     data_basename='CardiacModel',
     input_polydata=meshes,
-    mask_ids={1: 'ventricle', 2: 'atrium', 3: 'vessels'},
+    mask_ids=seg.taxonomy.all_labels(),
+    segmenter=seg,
     compute_normals=True
 )
 
@@ -463,98 +475,103 @@ Classes that inherit from `PhysioMotion4DBase` provide:
 - Class-based log filtering
 - Unified logging interface across the package
 
-## 📊 Experiments and Examples
+## Experiments and Examples
 
-The `experiments/` directory contains comprehensive Jupyter notebooks demonstrating the complete PhysioMotion4D pipeline:
+The `experiments/` directory contains research scripts that shaped the
+toolkit. They are `# %%` percent-cell Python scripts that can be run
+top-to-bottom (`python <script>.py`) or stepped through cell-by-cell in VS
+Code, Cursor, or any other editor with `# %%` cell-aware support. For the
+curated, supported user-facing entry points see `tutorials/` and
+`docs/tutorials.rst`.
 
-### 🫀 Heart-Gated CT (`experiments/Heart-GatedCT_To_USD/`)
+### Heart-Gated CT (`experiments/Heart-GatedCT_To_USD/`)
 
-Complete cardiac imaging workflow with step-by-step tutorials:
+Complete cardiac imaging workflow with step-by-step scripts:
 
-- **`0-download_and_convert_4d_to_3d.ipynb`**: Data preparation and 4D to 3D conversion
-- **`1-register_images.ipynb`**: Image registration between cardiac phases
-- **`2-generate_segmentation.ipynb`**: AI-based cardiac segmentation
-- **`3-transform_dynamic_and_static_contours.ipynb`**: Dynamic contour transformation
-- **`4-merge_dynamic_and_static_usd.ipynb`**: Final USD model creation and merging
+- **`0-download_and_convert_4d_to_3d.py`**: Data preparation and 4D to 3D conversion
+- **`1-register_images.py`**: Image registration between cardiac phases
+- **`2-generate_segmentation.py`**: AI-based cardiac segmentation
+- **`3-transform_dynamic_and_static_contours.py`**: Dynamic contour transformation
+- **`4-merge_dynamic_and_static_usd.py`**: Final USD model creation and merging
 
-**Sample Data**: The notebooks include instructions for downloading cardiac CT datasets from Slicer-Heart-CT.
+**Sample Data**: The scripts include instructions for downloading cardiac CT datasets from Slicer-Heart-CT.
 
-### 🫁 Lung-Gated CT (`experiments/Lung-GatedCT_To_USD/`)
+### Lung-Gated CT (`experiments/Lung-GatedCT_To_USD/`)
 
 Respiratory motion analysis using DirLab 4D-CT benchmark data:
 
-- **`0-register_dirlab_4dct.ipynb`**: Registration of respiratory phases
-- **`1-make_dirlab_models.ipynb`**: 3D model generation from lung segmentation
-- **`2-paint_dirlab_models.ipynb`**: USD material and visualization enhancement
+- **`0-register_dirlab_4dct.py`**: Registration of respiratory phases
+- **`1-make_dirlab_models.py`**: 3D model generation from lung segmentation
+- **`2-paint_dirlab_models.py`**: USD material and visualization enhancement
 
 **Sample Data**: Uses the standard DirLab 4D-CT benchmark datasets. DirLab data
 must be downloaded manually and placed under `data/DirLab-4DCT/`; see
 `data/README.md` for the expected layout.
 
-### 🎨 Colormap Visualization (`experiments/Colormap-VTK_To_USD/`)
+### Colormap Visualization (`experiments/Colormap-VTK_To_USD/`)
 
 Time-varying colormap rendering for scalar data visualization in Omniverse:
 
-- **`colormap_vtk_to_usd.ipynb`**: Convert VTK meshes with scalar data to USD with colormaps
+- **`colormap_vtk_to_usd.py`**: Convert VTK meshes with scalar data to USD with colormaps
 - Demonstrates plasma, viridis, rainbow, heat, coolwarm, grayscale, and custom colormaps
 
-### 🫀 Heart VTK Series (`experiments/Heart-VTKSeries_To_USD/`)
+### Heart VTK Series (`experiments/Heart-VTKSeries_To_USD/`)
 
 Direct VTK time series to USD conversion for cardiac data:
 
-- **`0-download_and_convert_4d_to_3d.ipynb`**: Data preparation
-- **`1-heart_vtkseries_to_usd.ipynb`**: VTK series to USD conversion
+- **`0-download_and_convert_4d_to_3d.py`**: Data preparation
+- **`1-heart_vtkseries_to_usd.py`**: VTK series to USD conversion
 
-### 🧠 Heart Create Statistical Model (`experiments/Heart-Create_Statistical_Model/`)
+### Heart Create Statistical Model (`experiments/Heart-Create_Statistical_Model/`)
 
 Create PCA statistical shape models from population meshes using the KCL Heart Model dataset:
 
-- **`1-input_meshes_to_input_surfaces.ipynb`**: Convert meshes to surfaces
-- **`2-input_surfaces_to_surfaces_aligned.ipynb`**: Align population meshes
-- **`3-registration_based_correspondence.ipynb`**: Compute point correspondences
-- **`4-surfaces_aligned_correspond_to_pca_inputs.ipynb`**: Prepare PCA inputs
-- **`5-compute_pca_model.ipynb`**: Compute PCA model using sklearn
+- **`1-input_meshes_to_input_surfaces.py`**: Convert meshes to surfaces
+- **`2-input_surfaces_to_surfaces_aligned.py`**: Align population meshes
+- **`3-registration_based_correspondence.py`**: Compute point correspondences
+- **`4-surfaces_aligned_correspond_to_pca_inputs.py`**: Prepare PCA inputs
+- **`5-compute_pca_model.py`**: Compute PCA model using sklearn
 
-**⚠️ Complete this experiment FIRST** before attempting `Heart-Statistical_Model_To_Patient`.
+**Complete this experiment FIRST** before attempting `Heart-Statistical_Model_To_Patient`.
 
-### 🧠 Heart Statistical Model to Patient (`experiments/Heart-Statistical_Model_To_Patient/`)
+### Heart Statistical Model to Patient (`experiments/Heart-Statistical_Model_To_Patient/`)
 
 Advanced registration between generic anatomical models and patient-specific data using PCA:
 
-- **`heart_model_to_model_icp_itk.ipynb`**: ICP registration for initial alignment
-- **`heart_model_to_model_registration_pca.ipynb`**: PCA-based statistical shape model registration
-- **`heart_model_to_patient.ipynb`**: Complete model-to-patient registration workflow
+- **`heart_model_to_model_icp_itk.py`**: ICP registration for initial alignment
+- **`heart_model_to_model_registration_pca.py`**: PCA-based statistical shape model registration
+- **`heart_model_to_patient.py`**: Complete model-to-patient registration workflow
 
 Uses the `WorkflowFitStatisticalModelToPatient` class for three-stage registration:
 1. ICP-based rough alignment
 2. Mask-to-mask deformable registration
 3. Optional PCA-constrained shape fitting
 
-### 🔬 4D CT Reconstruction (`experiments/Reconstruct4DCT/`)
+### 4D CT Reconstruction (`experiments/Reconstruct4DCT/`)
 
 Reconstruct 4D CT from sparse time samples using deformable registration:
 
-- **`reconstruct_4d_ct.ipynb`**: Temporal interpolation and 4D reconstruction
-- **`reconstruct_4d_ct_class.ipynb`**: Class-based reconstruction approach
+- **`reconstruct_4d_ct.py`**: Temporal interpolation and 4D reconstruction
+- **`reconstruct_4d_ct_class.py`**: Class-based reconstruction approach
 
-### 🫁 Vessel and Airway Segmentation (`experiments/Lung-VesselsAirways/`)
+### Vessel and Airway Segmentation (`experiments/Lung-VesselsAirways/`)
 
 Specialized deep learning for pulmonary vessel and airway segmentation:
 
-- **`0-GenData.ipynb`**: Training data generation for vessel segmentation models
+- **`0-GenData.py`**: Training data generation for vessel segmentation models
 - Includes trained ResNet18 models for vessel segmentation
 - Supporting branch structure test data
 
-### 🌊 Displacement Field Visualization (`experiments/DisplacementField_To_USD/`)
+### Displacement Field Visualization (`experiments/DisplacementField_To_USD/`)
 
 Convert image registration displacement fields to USD for advanced visualization:
 
-- **`displacement_field_to_usd.ipynb`**: Convert displacement fields to time-varying USD
+- **`displacement_field_to_usd.py`**: Convert displacement fields to time-varying USD
 - **`displacement_field_converter.py`**: DisplacementFieldToUSD class implementation
 - Integration with PhysicsNeMo for flow visualization in Omniverse
 - Supports streamlines, vector glyphs, and particle advection
 
-## 📥 Sample Data Sources
+## Sample Data Sources
 
 ### Cardiac Data
 - **Slicer-Heart-CT**: Cardiac gating examples from 3D Slicer
@@ -571,16 +588,13 @@ The Slicer-Heart sample can be downloaded directly from its public GitHub
 release:
 
 ```python
-from pathlib import Path
-from urllib.request import urlretrieve
+from physiomotion4d import DataDownloadTools
 
-data_dir = Path("data/Slicer-Heart-CT")
-data_dir.mkdir(parents=True, exist_ok=True)
-url = "https://github.com/SlicerHeart/SlicerHeart/releases/download/TestingData/TruncalValve_4DCT.seq.nrrd"
-urlretrieve(url, data_dir / "TruncalValve_4DCT.seq.nrrd")
+data_file = DataDownloadTools.DownloadSlicerHeartCTData("data/Slicer-Heart-CT")
+assert DataDownloadTools.VerifySlicerHeartCTData("data/Slicer-Heart-CT")
 ```
 
-## 🔧 Development
+## Development
 
 ### Code Quality Tools
 
@@ -795,7 +809,7 @@ without writing any code.
 /plan redesign the segmentation return type to use a dataclass instead of a tuple
 ```
 
-## 📖 Documentation
+## Additional Documentation
 
 The canonical documentation is published at
 https://project-monai.github.io/physiomotion4d/.
@@ -805,7 +819,7 @@ guides, contributing, testing, and troubleshooting. The `experiments/`
 directory records prior and ongoing experiments used to shape the toolkit; it
 is not the user-facing examples collection.
 
-## 🤝 Contributing
+## Contributing
 
 1. Fork the repository
 2. Create a feature branch (`git checkout -b feature/amazing-feature`)
@@ -816,11 +830,11 @@ is not the user-facing examples collection.
 
 See `docs/contributing.rst` for detailed contribution guidelines and IDE setup.
 
-## 📄 License
+## License
 
 This project is licensed under the Apache 2.0 License - see the LICENSE file for details.
 
-## 🙏 Acknowledgments
+## Acknowledgments
 
 - **NVIDIA Omniverse** team for USD format and visualization platform
 - **MONAI** community for medical imaging AI tools
@@ -828,13 +842,13 @@ This project is licensed under the Apache 2.0 License - see the LICENSE file for
 - **TotalSegmentator** team for segmentation models
 - **Icon Registration** team for deep learning registration methods
 
-## 📞 Support
+## Support
 
 - **Issues**: Report bugs and feature requests via GitHub Issues
 - **Discussions**: Join community discussions in GitHub Discussions
-- **Documentation**: Refer to docstrings and tutorial notebooks
+- **Documentation**: Refer to docstrings and tutorial scripts under `tutorials/`
 - **Examples**: Explore comprehensive examples in `experiments/` directory
 
 ---
 
-**Get started with the tutorial notebooks in `experiments/` to see PhysioMotion4D in action! 🚀**
+**Get started with the tutorial scripts under `tutorials/` to see PhysioMotion4D in action.**
diff --git a/data/KCL-Heart-Model/README.md b/data/KCL-Heart-Model/README.md
index eb67a4d..866fc3a 100644
--- a/data/KCL-Heart-Model/README.md
+++ b/data/KCL-Heart-Model/README.md
@@ -1,6 +1,6 @@
 # KCL Heart Model Dataset
 
-## ⚠️ Manual Download Required
+## Manual Download Required
 
 **This data is NOT automatically downloaded.** Users must manually download the dataset from Zenodo (see below).
 
@@ -33,7 +33,7 @@ This directory contains data from the King's College London (KCL) four-chamber h
 
 The KCL heart model data is publicly available through Zenodo:
 
-**🔗 [Virtual cohort of adult healthy four-chamber heart meshes from CT images](https://zenodo.org/records/4590294)**
+**[Virtual cohort of adult healthy four-chamber heart meshes from CT images](https://zenodo.org/records/4590294)**
 
 ### Dataset Details
 
@@ -52,7 +52,7 @@ If you use this dataset, please cite:
 
 This dataset is used in the PhysioMotion4D experiments for creating statistical shape models and patient-specific heart registrations.
 
-### 📓 Heart-Create_Statistical_Model Experiment
+### Heart-Create_Statistical_Model Experiment
 
 For a complete, step-by-step tutorial on creating a statistical heart shape model from this data, see:
 
@@ -63,7 +63,7 @@ This experiment demonstrates the complete workflow for:
 - Creating population-based shape models
 - Generating templates for patient-specific registration
 
-**⚠️ Start here first**: Complete the `Heart-Create_Statistical_Model` experiment before attempting the `Heart-Statistical_Model_To_Patient` experiment.
+**Start here first**: Complete the `Heart-Create_Statistical_Model` experiment before attempting the `Heart-Statistical_Model_To_Patient` experiment.
 
 ### Files in This Directory
 
diff --git a/data/README.md b/data/README.md
index c2ee1b5..ffc9aaf 100644
--- a/data/README.md
+++ b/data/README.md
@@ -2,7 +2,7 @@
 
 This directory contains sample datasets used for experiments, testing, and development of the PhysioMotion4D library. Each subdirectory contains a specific medical imaging dataset.
 
-## 📁 Directory Structure
+## Directory Structure
 
 ```
 data/
@@ -12,10 +12,10 @@ data/
 ├── CHOP-Valve4D/             # 4D valve models (MANUAL)
 ```
 
-## 📥 Data Download Methods
+## Data Download Methods
 
 ### Automatic Download (Only Slicer-Heart-CT)
-Only the **Slicer-Heart-CT** dataset can be automatically downloaded by running the appropriate notebook.
+Only the **Slicer-Heart-CT** dataset can be automatically downloaded by running the appropriate script.
 
 ### Manual Download (All Others)
 The following datasets must be **manually downloaded and preprocessed** by the user:
@@ -27,7 +27,7 @@ See individual dataset sections below for download instructions and preprocessin
 
 ---
 
-## 🫀 Slicer-Heart-CT ✅ AUTO-DOWNLOAD
+## Slicer-Heart-CT AUTO-DOWNLOAD
 
 ### Description
 4D cardiac CT dataset with temporal gating showing complete cardiac cycle motion. Pediatric cardiac CT with truncal valve visualization.
@@ -48,8 +48,10 @@ Data provided by Jolley Lab at CHOP (Children's Hospital of Philadelphia):
 
 **Automatic download** (recommended):
 ```python
-# The data will be automatically downloaded when running:
-# experiments/Heart-GatedCT_To_USD/0-download_and_convert_4d_to_3d.ipynb
+from physiomotion4d import DataDownloadTools
+
+DataDownloadTools.DownloadSlicerHeartCTData("data/Slicer-Heart-CT")
+assert DataDownloadTools.VerifySlicerHeartCTData("data/Slicer-Heart-CT")
 ```
 
 **Manual download** (alternative):
@@ -59,13 +61,28 @@ wget https://github.com/Slicer-Heart-CT/Slicer-Heart-CT/releases/download/Testin
 ```
 
 ### Usage
+- Primary dataset for tutorials
 - Primary dataset for `Heart-GatedCT_To_USD` experiments
 - Used in test suite (`tests/test_download_heart_data.py`)
 - Example data for cardiac motion visualization in Omniverse
 
+### Verification Helpers
+
+PhysioMotion4D exposes a small public utility for checking optional dataset
+layouts:
+
+```python
+from physiomotion4d import DataDownloadTools
+
+DataDownloadTools.VerifySlicerHeartCTData("data/Slicer-Heart-CT")
+DataDownloadTools.VerifyDirLab4DCTData("data/DirLab-4DCT")
+DataDownloadTools.VerifyKCLHeartModelData("data/KCL-Heart-Model")
+DataDownloadTools.VerifyCHOPValve4DData("data/CHOP-Valve4D")
+```
+
 ---
 
-## 🫁 DirLab-4DCT ⚠️ MANUAL DOWNLOAD
+## DirLab-4DCT MANUAL DOWNLOAD
 
 ### Description
 Benchmark dataset for 4D CT respiratory motion analysis. Contains 10 cases of lung CT scans at different respiratory phases with annotated landmark points for registration validation.
@@ -86,22 +103,22 @@ Data provided by the DIR-Lab at MD Anderson Cancer Center:
 
 ### Downloading the Data
 
-⚠️ **MANUAL DOWNLOAD REQUIRED**
+**MANUAL DOWNLOAD REQUIRED**
 
 Users must manually download and preprocess this dataset. Follow these steps:
 
 **Step 1: Manual Download**
 ```python
-# Using provided utilities in experiments notebooks
-# See: experiments/Lung-GatedCT_To_USD/0-register_dirlab_4dct.ipynb
-# The notebook includes download utilities but requires manual execution
+# Using provided utilities in experiment scripts.
+# See: experiments/Lung-GatedCT_To_USD/0-register_dirlab_4dct.py
+# The script includes download utilities but requires manual execution.
 ```
 
 **Step 2: User Preprocessing**
 Users are responsible for:
 - Downloading data from DIR-Lab website
 - Extracting and organizing files in the proper directory structure
-- Running preprocessing notebooks if needed
+- Running preprocessing scripts if needed
 
 ### Directory Structure
 ```
@@ -125,7 +142,7 @@ DirLab-4DCT/
 
 ---
 
-## 🧠 KCL-Heart-Model ⚠️ MANUAL DOWNLOAD
+## KCL-Heart-Model MANUAL DOWNLOAD
 
 ### Description
 Statistical shape model (SSM) of the human heart derived from cardiac imaging data. Includes principal component analysis (PCA) modes of shape variation.
@@ -158,7 +175,7 @@ Data from King's College London (KCL):
 
 ### Downloading the Data
 
-⚠️ **MANUAL DOWNLOAD REQUIRED**
+**MANUAL DOWNLOAD REQUIRED**
 
 Users must manually obtain and place this data:
 1. Obtain data from published research repositories or contact authors
@@ -177,7 +194,7 @@ Check the included PDFs (if available) for source information and proper citatio
 
 ---
 
-## 🫀 CHOP-Valve4D ⚠️ MANUAL DOWNLOAD
+## CHOP-Valve4D MANUAL DOWNLOAD
 
 ### Description
 Time-varying 4D valve reconstruction models showing valve motion over the cardiac cycle. These datasets represent dynamic valve geometries reconstructed from medical imaging data.
@@ -208,7 +225,7 @@ Data provided by Jolley Lab at CHOP (Children's Hospital of Philadelphia):
 
 ### Downloading the Data
 
-⚠️ **MANUAL DOWNLOAD REQUIRED**
+**MANUAL DOWNLOAD REQUIRED**
 
 **Availability**: This dataset will soon be publicly available for download from the **FEBio website** under the **Creative Commons Attribution (CC-BY) license**.
 
@@ -223,7 +240,7 @@ Data provided by Jolley Lab at CHOP (Children's Hospital of Philadelphia):
 4. Organize by valve type in subdirectories (e.g., `Alterra/`, `TPV25/`)
 
 ### Usage
-- Time-series VTK to USD conversion (`experiments/convert_vtk_to_usd_lib/valve4d_time_series.ipynb`)
+- Time-series VTK to USD conversion (`experiments/Convert_VTK_To_USD/`)
 - 4D valve motion visualization in NVIDIA Omniverse
 - Temporal cardiac mechanics analysis
 - Valve dynamics studies and surgical planning
@@ -234,7 +251,7 @@ Data provided by Jolley Lab at CHOP (Children's Hospital of Philadelphia):
 
 ---
 
-## 📝 Data Usage Guidelines
+## Data Usage Guidelines
 
 ### For Testing
 - Tests automatically use cached data when available
@@ -253,26 +270,27 @@ Data provided by Jolley Lab at CHOP (Children's Hospital of Philadelphia):
 
 ---
 
-## 🔒 Data Access and Licensing
+## Data Access and Licensing
 
-- **Slicer-Heart-CT** ✅: Public release from GitHub (auto-download available)
-- **DirLab-4DCT** ⚠️: Public benchmark dataset (manual download required, may require registration)
-- **KCL-Heart-Model** ⚠️: Requires manual download from research repositories
-- **CHOP-Valve4D** ⚠️: Soon available from FEBio website under CC-BY license (manual download)
+- **Slicer-Heart-CT** : Public release from GitHub (auto-download available)
+- **DirLab-4DCT** : Public benchmark dataset (manual download required, may require registration)
+- **KCL-Heart-Model** : Requires manual download from research repositories
+- **CHOP-Valve4D** : Soon available from FEBio website under CC-BY license (manual download)
 
-⚠️ **Important**: Always cite the original data sources in publications and respect any usage restrictions.
+**Important**: Always cite the original data sources in publications and respect any usage restrictions.
 
 ### Summary of Download Methods
+
 | Dataset         | Auto-Download | Manual Required | License         | Source              | Used in Tests            |
 | --------------- | ------------- | --------------- | --------------- | ------------------- | ------------------------ |
-| Slicer-Heart-CT | ✅ Yes         | No              | Public          | GitHub              | Yes                      |
-| DirLab-4DCT     | ❌ No          | Yes             | Public/Academic | DIR-Lab             | No                       |
-| KCL-Heart-Model | ❌ No          | Yes             | Check citation  | Zenodo/KCL          | Yes (skipped if missing) |
-| CHOP-Valve4D    | ❌ No          | Yes             | CC-BY           | FEBio (coming soon) | No                       |
+| Slicer-Heart-CT | Yes         | No              | Public          | GitHub              | Yes                      |
+| DirLab-4DCT     | No          | Yes             | Public/Academic | DIR-Lab             | No                       |
+| KCL-Heart-Model | No          | Yes             | Check citation  | Zenodo/KCL          | Yes (skipped if missing) |
+| CHOP-Valve4D    | No          | Yes             | CC-BY           | FEBio (coming soon) | No                       |
 
 ---
 
-## 📚 References
+## References
 
 ### Slicer-Heart-CT
 - Jolley Lab: https://www.linkedin.com/company/jolleylab
@@ -295,7 +313,7 @@ Data provided by Jolley Lab at CHOP (Children's Hospital of Philadelphia):
 
 ---
 
-## 💡 Tips
+## Tips
 
 1. **Storage**: Ensure adequate disk space (~10-20GB for all datasets)
 2. **Download Time**: Initial downloads can be slow; be patient
diff --git a/data/Slicer-Heart-CT/download_and_convert.py b/data/Slicer-Heart-CT/download_and_convert.py
index de37c12..ef52f74 100644
--- a/data/Slicer-Heart-CT/download_and_convert.py
+++ b/data/Slicer-Heart-CT/download_and_convert.py
@@ -2,9 +2,9 @@
 # %%
 import os
 import shutil
-import urllib.request
 
 from physiomotion4d.convert_nrrd_4d_to_3d import ConvertNRRD4DTo3D
+from physiomotion4d.data_download_tools import DataDownloadTools
 
 # %%
 data_dir = "."
@@ -16,17 +16,12 @@
 if not os.path.exists(output_dir):
     os.makedirs(output_dir, exist_ok=True)
 
-# %%
-input_image_url = "https://github.com/SlicerHeart/SlicerHeart/releases/download/TestingData/TruncalValve_4DCT.seq.nrrd"
-input_image_filename = os.path.join(data_dir, "TruncalValve_4DCT.seq.nrrd")
-
-if not os.path.exists(input_image_filename):
-    urllib.request.urlretrieve(input_image_url, input_image_filename)
+input_image_filename = DataDownloadTools.DownloadSlicerHeartCTData(data_dir)
 
 # %%
 conv = ConvertNRRD4DTo3D()
-conv.load_nrrd_4d(f"{data_dir}/TruncalValve_4DCT.seq.nrrd")
-conv.save_3d_images(f"{output_dir}/slice")
+conv.load_nrrd_4d(str(input_image_filename))
+conv.save_3d_images(output_dir, "slice")
 
 # Save the mid-stroke slice as the fixed/reference image
 shutil.copyfile(f"{output_dir}/slice_007.mha", f"{output_dir}/slice_fixed.mha")
diff --git a/data/test/.gitignore b/data/test/.gitignore
index 1850c69..d1c4ec7 100644
--- a/data/test/.gitignore
+++ b/data/test/.gitignore
@@ -1,3 +1,2 @@
-*.hdf
-*.mha
-*.nrrd
+slicer_heart
+slicer_heart_small
diff --git a/docs/API_MAP.md b/docs/API_MAP.md
index fd4d2e0..7624f31 100644
--- a/docs/API_MAP.md
+++ b/docs/API_MAP.md
@@ -26,21 +26,6 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 
 - `def generate_pc_variation(pc_index, std_dev_multiplier=3.0)` (line 155): Generate shape variations along a principal component.
 
-## experiments/Heart-GatedCT_To_USD/3-transform_dynamic_and_static_contours.py
-
-- `def transform_contours(contours, transform_filenames, frame_indices, base_name, output_dir)` (line 31)
-- `def convert_contours(base_name, output_dir, project_name, compute_normals=False)` (line 49)
-
-## experiments/Lung-GatedCT_To_USD/0-register_dirlab_4dct.py
-
-- `def dilate_mask(mask, dilation)` (line 30)
-- `def register_image(fixed_image, fixed_mask, moving_image, moving_mask, case_name, image_num, mask_name, output_dir)` (line 39): Register a moving image to a fixed image using a mask.
-
-## experiments/Lung-GatedCT_To_USD/1-make_dirlab_models.py
-
-- `def transform_contours_list(contours, case_name, mask_name, output_dir)` (line 22): Transform a list of contours to a list of transformed contours.
-- `def make_dirlab_models(output_dir, label, case_name, base_timepoint, all_labelmap_arr, all_mask_ids, con_tools)` (line 42): Make DirLab models for a list of cases.
-
 ## experiments/Lung-GatedCT_To_USD/data_dirlab_4d_ct.py
 
 - **class DataDirLab4DCT** (line 10): This class is used to store the data for the DirLab 4DCT dataset.
@@ -63,6 +48,20 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 
 - `def register_slices(reg_tool, reg_tool_name, fixed_image, images, files_indx, reference_image_num, reference_image_reg_use_identity, portion_of_prior_to_use=0.0)` (line 62)
 
+## src/physiomotion4d/anatomy_taxonomy.py
+
+- **class AnatomyGroup** (line 26): One named anatomy group together with the organ labels it contains.
+- **class AnatomyTaxonomy** (line 38): Mapping of anatomical groups to the organs each group contains.
+  - `def __init__(self)` (line 58)
+  - `def add_group(self, name)` (line 61): Ensure a group exists and return it.
+  - `def add_organ(self, group, label_id, organ_name)` (line 74): Add one organ label to the named group.
+  - `def group_names(self)` (line 104): Return group names in the order they were first added.
+  - `def labels_in_group(self, group)` (line 108): Return ``{label_id: organ_name}`` for *group*; empty dict if absent.
+  - `def all_labels(self)` (line 113): Return the union of every group's organs as a single id→name dict.
+  - `def group_for_label(self, label_name)` (line 120): Return the group containing *label_name*.
+  - `def group_for_id(self, label_id)` (line 132): Return the group containing *label_id*; :data:`OTHER_GROUP` if absent.
+  - `def fill_other_group(self, id_range=range(1, 256), name_template='other_{id}')` (line 139): Populate the ``other`` group with any ids not already claimed.
+
 ## src/physiomotion4d/cli/convert_ct_to_vtk.py
 
 - `def main()` (line 26): CLI entry point for CT to VTK conversion.
@@ -73,7 +72,7 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 
 ## src/physiomotion4d/cli/convert_vtk_to_usd.py
 
-- `def main()` (line 28): Command-line interface for VTK to USD conversion.
+- `def main()` (line 22): Command-line interface for VTK to USD conversion.
 
 ## src/physiomotion4d/cli/create_statistical_model.py
 
@@ -105,24 +104,34 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 
 ## src/physiomotion4d/convert_nrrd_4d_to_3d.py
 
-- **class ConvertNRRD4DTo3D** (line 11)
-  - `def __init__(self, log_level=logging.INFO)` (line 12): Initialize the NRRD 4D to 3D converter.
-  - `def load_nrrd_3d(self, filenames)` (line 22)
-  - `def load_nrrd_4d(self, filename)` (line 28)
-  - `def get_3d_image(self, index)` (line 62)
-  - `def get_number_of_3d_images(self)` (line 65)
-  - `def save_3d_images(self, basename)` (line 68)
+- **class ConvertNRRD4DTo3D** (line 13)
+  - `def __init__(self, log_level=logging.INFO)` (line 14): Initialize the NRRD 4D to 3D converter.
+  - `def load_nrrd_3d(self, filenames)` (line 24)
+  - `def load_nrrd_4d(self, filename)` (line 30)
+  - `def get_3d_image(self, index)` (line 64)
+  - `def get_number_of_3d_images(self)` (line 67)
+  - `def save_3d_images(self, directory, basename)` (line 70)
 
 ## src/physiomotion4d/convert_vtk_to_usd.py
 
-- **class ConvertVTKToUSD** (line 43): Advanced VTK to USD converter with colormap and anatomical labeling support.
-  - `def __init__(self, data_basename, input_polydata, mask_ids=None, compute_normals=False, convert_to_surface=True, times_per_second=24.0, separate_by='none', solid_color=(0.8, 0.8, 0.8), log_level=logging.INFO)` (line 73): Initialize converter.
-  - `def from_files(cls, data_basename, vtk_files, *, extract_surface=True, separate_by='none', times_per_second=24.0, solid_color=(0.8, 0.8, 0.8), time_codes=None, static_merge=False, mask_ids=None, log_level=logging.INFO)` (line 144): Create a converter by loading VTK files from disk.
-  - `def supports_mesh_type(self, mesh)` (line 242): Check if mesh type is supported for conversion.
-  - `def inspect_file(cls, vtk_file, *, extract_surface=True)` (line 271): Summarize a VTK file using the same low-level reader as conversion.
-  - `def list_available_arrays(self)` (line 343): List all point data arrays available across all time steps.
-  - `def set_colormap(self, color_by_array=None, colormap='plasma', intensity_range=None)` (line 389): Configure colormap for visualization.
-  - `def convert(self, output_usd_file, convert_to_surface=None, compute_normals=None)` (line 423): Convert VTK meshes to USD.
+- **class ConvertVTKToUSD** (line 45): Advanced VTK to USD converter with colormap and anatomical labeling support.
+  - `def __init__(self, data_basename, input_polydata, mask_ids=None, compute_normals=False, convert_to_surface=True, times_per_second=24.0, separate_by='none', solid_color=(0.8, 0.8, 0.8), segmenter=None, log_level=logging.INFO)` (line 75): Initialize converter.
+  - `def from_files(cls, data_basename, vtk_files, *, extract_surface=True, separate_by='none', times_per_second=24.0, solid_color=(0.8, 0.8, 0.8), time_codes=None, static_merge=False, mask_ids=None, segmenter=None, log_level=logging.INFO)` (line 154): Create a converter by loading VTK files from disk.
+  - `def supports_mesh_type(self, mesh)` (line 256): Check if mesh type is supported for conversion.
+  - `def inspect_file(cls, vtk_file, *, extract_surface=True)` (line 285): Summarize a VTK file using the same low-level reader as conversion.
+  - `def list_available_arrays(self)` (line 357): List all point data arrays available across all time steps.
+  - `def set_colormap(self, color_by_array=None, colormap='plasma', intensity_range=None)` (line 403): Configure colormap for visualization.
+  - `def compute_von_mises_stress(self, stress_array_name='stress', output_name='von_mises_stress')` (line 437): Add a scalar von Mises stress array derived from a 9-component
+  - `def convert(self, output_usd_file, convert_to_surface=None, compute_normals=None)` (line 538): Convert VTK meshes to USD.
+
+## src/physiomotion4d/data_download_tools.py
+
+- **class DataDownloadTools** (line 20): Download and verify optional PhysioMotion4D example datasets.
+  - `def DownloadSlicerHeartCTData(dirname)` (line 30): Download the Slicer-Heart-CT 4-D CT sample into ``dirname``.
+  - `def VerifySlicerHeartCTData(dirname)` (line 80): Return True when Slicer-Heart-CT has the expected 4-D CT file.
+  - `def VerifyCHOPValve4DData(dirname)` (line 85): Return True when CHOP-Valve4D files referenced by the repo exist.
+  - `def VerifyDirLab4DCTData(dirname)` (line 107): Return True when a supported DirLab-4DCT case layout exists.
+  - `def VerifyKCLHeartModelData(dirname)` (line 122): Return True when KCL-Heart-Model has its expected mesh inputs.
 
 ## src/physiomotion4d/image_tools.py
 
@@ -135,10 +144,6 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
   - `def convert_array_to_image_of_vectors(self, arr_data, reference_image, ptype=itk.D)` (line 218): Convert a numpy array to an ITK image of vector type.
   - `def flip_image(self, in_image, in_mask=None, flip_x=False, flip_y=False, flip_z=False, flip_and_make_identity=False)` (line 249): Flip the image and mask.
 
-## src/physiomotion4d/notebook_utils.py
-
-- `def running_as_test()` (line 11): True when the notebook is run as a test (e.g. by pytest experiment tests).
-
 ## src/physiomotion4d/physiomotion4d_base.py
 
 - **class ClassNameFilter** (line 38): Filter to show logs only from specific class names.
@@ -160,14 +165,14 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 
 ## src/physiomotion4d/register_images_ants.py
 
-- **class RegisterImagesANTs** (line 25): ANTs-based deformable image registration implementation.
-  - `def __init__(self, log_level=logging.INFO)` (line 71): Initialize the ANTs image registration class.
-  - `def set_number_of_iterations(self, number_of_iterations)` (line 86): Set the number of iterations for ANTs registration.
-  - `def set_transform_type(self, transform_type)` (line 95): Set the type of transform to use for registration.
-  - `def set_metric(self, metric)` (line 107): Set the similarity metric to use for registration.
-  - `def itk_affine_transform_to_ants_transform(self, itk_tfm)` (line 317): Convert ITK affine/rigid transform to ANTs affine transform.
-  - `def itk_transform_to_antsfile(self, itk_tfm, reference_image, output_filename)` (line 410): Convert ITK transform to ANTs transform file.
-  - `def registration_method(self, moving_image, moving_mask=None, moving_labelmap=None, moving_image_pre=None, initial_forward_transform=None)` (line 510): Register moving image to fixed image using ANTs registration algorithm.
+- **class RegisterImagesANTs** (line 24): ANTs-based deformable image registration implementation.
+  - `def __init__(self, log_level=logging.INFO)` (line 70): Initialize the ANTs image registration class.
+  - `def set_number_of_iterations(self, number_of_iterations)` (line 85): Set the number of iterations for ANTs registration.
+  - `def set_transform_type(self, transform_type)` (line 94): Set the type of transform to use for registration.
+  - `def set_metric(self, metric)` (line 106): Set the similarity metric to use for registration.
+  - `def itk_affine_transform_to_ants_transform(self, itk_tfm)` (line 316): Convert ITK affine/rigid transform to ANTs affine transform.
+  - `def itk_transform_to_antsfile(self, itk_tfm, reference_image, output_filename)` (line 409): Convert ITK transform to ANTs transform file.
+  - `def registration_method(self, moving_image, moving_mask=None, moving_labelmap=None, moving_image_pre=None, initial_forward_transform=None)` (line 509): Register moving image to fixed image using ANTs registration algorithm.
 
 ## src/physiomotion4d/register_images_base.py
 
@@ -254,102 +259,105 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 
 ## src/physiomotion4d/segment_anatomy_base.py
 
-- **class SegmentAnatomyBase** (line 18): Base class for anatomy segmentation that provides common functionality for
-  - `def __init__(self, log_level=logging.INFO)` (line 45): Initialize the SegmentAnatomyBase class.
-  - `def set_other_and_all_mask_ids(self)` (line 78): Set the other mask IDs and consolidate all mask ID dictionaries.
-  - `def set_target_spacing(self, target_spacing)` (line 111): Set the target isotropic spacing for image resampling.
-  - `def preprocess_input(self, input_image)` (line 123): Preprocess the input image for segmentation.
-  - `def postprocess_labelmap(self, labelmap_image, input_image)` (line 245): Resample the labelmap to match the input image spacing.
-  - `def segment_connected_component(self, preprocessed_image, labelmap_image, lower_threshold, upper_threshold, labelmap_ids=None, mask_id=0, use_mid_slice=True, hole_fill=2)` (line 341): Segment connected components based on intensity thresholding.
-  - `def segment_contrast_agent(self, preprocessed_image, labelmap_image)` (line 448): Include contrast-enhanced blood in the labelmap.
-  - `def create_anatomy_group_masks(self, labelmap_image)` (line 490): Create binary masks for different anatomical groups from the labelmap.
-  - `def segmentation_method(self, preprocessed_image)` (line 576): Abstract method for image segmentation - must be implemented by subclasses.
-  - `def dilate_mask(self, mask, dilation)` (line 598): Dilate a binary mask using morphological operations.
-  - `def segment(self, input_image, contrast_enhanced_study=False)` (line 621): Perform complete chest CT segmentation.
+- **class SegmentAnatomyBase** (line 19): Base class for anatomy segmentation that provides common functionality for
+  - `def __init__(self, log_level=logging.INFO)` (line 50): Initialize the SegmentAnatomyBase class.
+  - `def label_to_type(self, label_name)` (line 92): Return the anatomy group ('heart', 'lung', etc.) for a label name.
+  - `def set_target_spacing(self, target_spacing)` (line 108): Set the target isotropic spacing for image resampling.
+  - `def preprocess_input(self, input_image)` (line 120): Preprocess the input image for segmentation.
+  - `def postprocess_labelmap(self, labelmap_image, input_image)` (line 242): Resample the labelmap to match the input image spacing.
+  - `def segment_connected_component(self, preprocessed_image, labelmap_image, lower_threshold, upper_threshold, labelmap_ids=None, mask_id=0, use_mid_slice=True, hole_fill=2)` (line 338): Segment connected components based on intensity thresholding.
+  - `def segment_contrast_agent(self, preprocessed_image, labelmap_image)` (line 445): Include contrast-enhanced blood in the labelmap.
+  - `def create_anatomy_group_masks(self, labelmap_image)` (line 492): Create binary masks for different anatomical groups from the labelmap.
+  - `def segmentation_method(self, preprocessed_image)` (line 538): Abstract method for image segmentation - must be implemented by subclasses.
+  - `def dilate_mask(self, mask, dilation)` (line 560): Dilate a binary mask using morphological operations.
+  - `def segment(self, input_image, contrast_enhanced_study=False)` (line 583): Perform complete chest CT segmentation.
 
 ## src/physiomotion4d/segment_chest_total_segmentator.py
 
 - **class SegmentChestTotalSegmentator** (line 21): Chest CT segmentation using TotalSegmentator deep learning model.
-  - `def __init__(self, log_level=logging.INFO)` (line 57): Initialize the TotalSegmentator-based chest segmentation.
-  - `def segmentation_method(self, preprocessed_image)` (line 199): Run TotalSegmentator on the preprocessed image and return result.
+  - `def __init__(self, log_level=logging.INFO)` (line 50): Initialize the TotalSegmentator-based chest segmentation.
+  - `def segmentation_method(self, preprocessed_image)` (line 209): Run TotalSegmentator on the preprocessed image and return result.
 
 ## src/physiomotion4d/segment_heart_simpleware.py
 
 - **class SegmentHeartSimpleware** (line 23): Heart CT segmentation using Simpleware Medical's ASCardio module.
-  - `def __init__(self, log_level=logging.INFO)` (line 56): Initialize the Simpleware Medical based heart segmentation.
+  - `def __init__(self, log_level=logging.INFO)` (line 58): Initialize the Simpleware Medical based heart segmentation.
   - `def set_trim_mask_to_essentials(self, trim_mask)` (line 118): Set whether to trim mask to common and critical structures.
   - `def set_simpleware_executable_path(self, path)` (line 126): Set the path to the Simpleware Medical console executable.
   - `def segmentation_method(self, preprocessed_image)` (line 139): Run Simpleware Medical ASCardio segmentation on the preprocessed image.
-  - `def get_landmarks(self)` (line 327): Get the landmarks.
-  - `def trim_mask_to_essentials(self, labelmap_image)` (line 331): Trim mask to essentials.
+  - `def get_landmarks(self)` (line 339): Get the landmarks.
+  - `def trim_mask_to_essentials(self, labelmap_image)` (line 343): Trim mask to essentials.
 
 ## src/physiomotion4d/test_tools.py
 
-- `def set_create_baseline_if_missing(value)` (line 28): Set whether to create baseline files when missing (used by pytest conftest).
-- **class TestTools** (line 34): Utilities for pytest image comparison: baseline directory, result directory,
-  - `def __init__(self, results_dir, baselines_dir, class_name, *, results_output_dir=None, log_level=logging.INFO)` (line 44): Initialize test helpers.
-  - `def image_pass_fail_and_pixels_above_tolerance(self)` (line 95): Return (pass, value) for number of pixels above tolerance from the most
-  - `def image_pass_fail_and_total_absolute_error(self)` (line 110): Return (pass, value) for total absolute error from the most recent
-  - `def image_difference(self)` (line 125): Return the difference image (itk.Image) from the most recent
-  - `def transform_pass_fail_and_number_of_values_above_tolerance(self)` (line 132): Return (pass, value) for number of values above tolerance from the most recent compare_result_to_baseline_transform call.
-  - `def transform_pass_fail_and_total_absolute_error(self)` (line 148): Return (pass, value) for total absolute error from the most recent compare_result_to_baseline_transform call.
-  - `def transform_difference(self)` (line 162): Return the difference transform (itk.Transform) from the most recent compare_result_to_baseline_transform call.
-  - `def write_result_image(self, image, filename)` (line 168): Write the image to the configured result artifact directory.
-  - `def write_result_transform(self, transform, filename)` (line 172): Write the transform to the configured result artifact directory.
-  - `def compare_result_to_baseline_transform(self, filename, *, per_value_absolute_error_tol=0.0, max_number_of_values_above_tol=0, total_absolute_error_tol=0.0)` (line 178): Compare the transform to the baseline transform.
-  - `def compare_result_to_baseline_image(self, filename, *, per_pixel_absolute_error_tol=0.0, max_number_of_pixels_above_tol=0, total_absolute_error_tol=0.0)` (line 256): Load a 3D result image and a 3D baseline image (.mha), compare the full
-  - `def save_screenshot_mesh(self, mesh, filename, *, camera_position='iso', window_size=(800, 600), color='pink', opacity=0.9)` (line 363): Render a PyVista mesh off-screen and save a PNG.
-  - `def save_screenshot_image_slice(self, image, filename, *, axis=0, slice_fraction=0.5, colormap='gray', vmin=None, vmax=None, overlay_mask=None, overlay_alpha=0.4)` (line 412): Extract one slice from an ITK image and save a PNG via matplotlib.
+- `def set_create_baseline_if_missing(value)` (line 29): Set whether to create baseline files when missing (used by pytest conftest).
+- **class TestTools** (line 35): Utilities for pytest image comparison: baseline directory, result directory,
+  - `def __init__(self, class_name, results_dir=None, baselines_dir=None, *, log_level=logging.INFO)` (line 45): Initialize test helpers.
+  - `def running_as_test()` (line 103): True when the script is run as a test (e.g. by pytest experiment tests).
+  - `def image_pass_fail_and_pixels_above_tolerance(self)` (line 121): Return (pass, value) for number of pixels above tolerance from the most
+  - `def image_pass_fail_and_total_absolute_error(self)` (line 136): Return (pass, value) for total absolute error from the most recent
+  - `def image_difference(self)` (line 151): Return the difference image (itk.Image) from the most recent
+  - `def transform_pass_fail_and_number_of_values_above_tolerance(self)` (line 158): Return (pass, value) for number of values above tolerance from the most recent compare_result_to_baseline_transform call.
+  - `def transform_pass_fail_and_total_absolute_error(self)` (line 174): Return (pass, value) for total absolute error from the most recent compare_result_to_baseline_transform call.
+  - `def transform_difference(self)` (line 188): Return the difference transform (itk.Transform) from the most recent compare_result_to_baseline_transform call.
+  - `def write_result_image(self, image, filename)` (line 194): Write the image to the configured result artifact directory.
+  - `def write_result_transform(self, transform, filename)` (line 198): Write the transform to the configured result artifact directory.
+  - `def compare_result_to_baseline_transform(self, filename, *, per_value_absolute_error_tol=0.0, max_number_of_values_above_tol=0, total_absolute_error_tol=0.0)` (line 204): Compare the transform to the baseline transform.
+  - `def compare_result_to_baseline_image(self, filename, *, per_pixel_absolute_error_tol=0.0, max_number_of_pixels_above_tol=0, total_absolute_error_tol=0.0)` (line 282): Load a 3D result image and a 3D baseline image (.mha), compare the full
+  - `def save_screenshot_mesh(self, mesh, filename, *, camera_position='iso', window_size=(800, 600), color='pink', opacity=0.9)` (line 389): Render a PyVista mesh off-screen and save a PNG.
+  - `def save_screenshot_openusd(self, usd_file, filename, *, prim_path='/World', time_code=None)` (line 438): Render USD mesh geometry off-screen and save a PNG.
+  - `def save_screenshot_image_slice(self, image, filename, *, axis=0, slice_fraction=0.5, colormap='gray', vmin=None, vmax=None, overlay_mask=None, overlay_alpha=0.4)` (line 507): Extract one slice from an ITK image and save a PNG via matplotlib.
 
 ## src/physiomotion4d/transform_tools.py
 
-- **class TransformTools** (line 39): Utilities for transforming and manipulating ITK transforms.
-  - `def __init__(self, log_level=logging.INFO)` (line 72): Initialize the TransformTools class.
-  - `def combine_displacement_field_transforms(self, tfm1, tfm2, reference_image, tfm1_weight=1.0, tfm2_weight=1.0, mode='compose', tfm1_blur_sigma=0.0, tfm2_blur_sigma=0.0)` (line 80): Compose two displacement field transforms.
-  - `def convert_transform_to_displacement_field(self, tfm, reference_image, np_component_type=np.float64, use_reference_image_as_mask=False)` (line 164): Generate a dense deformation field from an ITK transform.
-  - `def convert_transform_to_displacement_field_transform(self, tfm, reference_image)` (line 252): Convert an ITK transform to a displacement field transform.
-  - `def invert_displacement_field_transform(self, tfm)` (line 269): Invert a displacement field transform.
-  - `def transform_pvcontour(self, contour, tfm, with_deformation_magnitude=False)` (line 291): Transform PyVista contour meshes using an ITK transform.
-  - `def transform_dataset(self, mesh, tfm, with_deformation_magnitude=False)` (line 335): Transform a PyVista dataset while preserving mesh topology and data arrays.
-  - `def transform_image(self, img, tfm, reference_image, interpolation_method='linear')` (line 382): Transform an ITK image using a specified transform and interpolation.
-  - `def convert_vtk_matrix_to_itk_transform(self, vtk_mat)` (line 459): Convert a VTK matrix to an ITK transform.
-  - `def smooth_transform(self, tfm, sigma, reference_image)` (line 494): Smooth a transform using Gaussian filtering to reduce noise.
-  - `def combine_transforms_with_masks(self, transform1, transform2, mask1, mask2, reference_image, max_iter=10, jacobian_threshold=0.1)` (line 553): Combine two transforms using spatial masks with folding correction.
-  - `def compute_jacobian_determinant_from_field(self, field)` (line 643): Compute Jacobian determinant of a displacement field.
-  - `def detect_folding_in_field(self, jacobian_det, threshold=0.1)` (line 672): Detect spatial folding in a transform.
-  - `def reduce_folding_in_field(self, field, jacobian_det, reduction_factor=0.8, threshold=0.1)` (line 696): Reduce folding by scaling displacement field in problematic regions.
-  - `def generate_grid_image(self, reference_image, grid_size=60, line_width=3)` (line 742): Generate a grid image.
-  - `def convert_field_to_grid_visualization(self, tfm, reference_image, grid_size=60, line_width=3)` (line 775): Generate a visual deformation grid for transform visualization.
-  - `def convert_itk_transform_to_usd_visualization(self, tfm, reference_image, output_filename, visualization_type='arrows', subsample_factor=4, arrow_scale=1.0, magnitude_threshold=0.0)` (line 811): Convert an ITK transform to a USD visualization for NVIDIA Omniverse.
+- **class TransformTools** (line 41): Utilities for transforming and manipulating ITK transforms.
+  - `def __init__(self, log_level=logging.INFO)` (line 74): Initialize the TransformTools class.
+  - `def combine_displacement_field_transforms(self, tfm1, tfm2, reference_image, tfm1_weight=1.0, tfm2_weight=1.0, mode='compose', tfm1_blur_sigma=0.0, tfm2_blur_sigma=0.0)` (line 82): Compose two displacement field transforms.
+  - `def convert_transform_to_displacement_field(self, tfm, reference_image, np_component_type=np.float64, use_reference_image_as_mask=False)` (line 166): Generate a dense deformation field from an ITK transform.
+  - `def convert_transform_to_displacement_field_transform(self, tfm, reference_image)` (line 254): Convert an ITK transform to a displacement field transform.
+  - `def invert_displacement_field_transform(self, tfm)` (line 271): Invert a displacement field transform.
+  - `def transform_pvcontour(self, contour, tfm, with_deformation_magnitude=False)` (line 293): Transform PyVista contour meshes using an ITK transform.
+  - `def transform_dataset(self, mesh, tfm, with_deformation_magnitude=False)` (line 337): Transform a PyVista dataset while preserving mesh topology and data arrays.
+  - `def transform_image(self, img, tfm, reference_image, interpolation_method='linear')` (line 384): Transform an ITK image using a specified transform and interpolation.
+  - `def convert_vtk_matrix_to_itk_transform(self, vtk_mat)` (line 461): Convert a VTK matrix to an ITK transform.
+  - `def smooth_transform(self, tfm, sigma, reference_image)` (line 496): Smooth a transform using Gaussian filtering to reduce noise.
+  - `def combine_transforms_with_masks(self, transform1, transform2, mask1, mask2, reference_image, max_iter=10, jacobian_threshold=0.1)` (line 555): Combine two transforms using spatial masks with folding correction.
+  - `def compute_jacobian_determinant_from_field(self, field)` (line 645): Compute Jacobian determinant of a displacement field.
+  - `def detect_folding_in_field(self, jacobian_det, threshold=0.1)` (line 674): Detect spatial folding in a transform.
+  - `def reduce_folding_in_field(self, field, jacobian_det, reduction_factor=0.8, threshold=0.1)` (line 698): Reduce folding by scaling displacement field in problematic regions.
+  - `def generate_grid_image(self, reference_image, grid_size=60, line_width=3)` (line 744): Generate a grid image.
+  - `def convert_field_to_grid_visualization(self, tfm, reference_image, grid_size=60, line_width=3)` (line 777): Generate a visual deformation grid for transform visualization.
+  - `def convert_itk_transform_to_usd_visualization(self, tfm, reference_image, output_filename, visualization_type='arrows', subsample_factor=4, arrow_scale=1.0, magnitude_threshold=0.0)` (line 813): Convert an ITK transform to a USD visualization for NVIDIA Omniverse.
 
 ## src/physiomotion4d/usd_anatomy_tools.py
 
-- **class USDAnatomyTools** (line 14): This class is used to enhance the appearance of anatomy meshes in a USD
-  - `def __init__(self, stage, log_level=logging.INFO)` (line 20): Initialize USDAnatomyTools.
-  - `def get_anatomy_types(self)` (line 195): Return list of supported anatomy type names for apply_anatomy_material_to_mesh.
-  - `def get_anatomy_diffuse_color(self, anatomy_type)` (line 199): Return the diffuse reflection RGB color for the given anatomy type.
-  - `def apply_anatomy_material_to_mesh(self, mesh_path, anatomy_type)` (line 226): Apply an anatomic OmniSurface material to a single mesh prim by type.
-  - `def apply_anatomy_material_to_prim(self, prim, material_params)` (line 250): Corrected material application with Omniverse-specific fixes
-  - `def enhance_meshes(self, segmentator)` (line 321): Find and enhance all heart meshes
+- **class USDAnatomyTools** (line 207): Apply OmniSurface materials to anatomy mesh prims in a USD stage.
+  - `def __init__(self, stage, log_level=logging.INFO)` (line 216): Initialize USDAnatomyTools.
+  - `def get_anatomy_types(self)` (line 233): Return list of registered render-param keys (groups + organ overrides).
+  - `def get_anatomy_diffuse_color(self, anatomy_type)` (line 237): Return the diffuse reflection RGB color for the given group/organ.
+  - `def apply_anatomy_material_to_mesh(self, mesh_path, anatomy_type)` (line 264): Apply an anatomic OmniSurface material to a single mesh prim by type.
+  - `def apply_anatomy_material_to_prim(self, prim, material_params)` (line 288): Corrected material application with Omniverse-specific fixes
+  - `def enhance_meshes(self, segmentator)` (line 359): Apply per-organ OmniSurface materials to every matching mesh prim.
 
 ## src/physiomotion4d/usd_tools.py
 
-- **class USDTools** (line 25): Utilities for manipulating Universal Scene Description (USD) files.
-  - `def __init__(self, log_level=logging.INFO)` (line 61): Initialize the USDTools class.
-  - `def get_subtree_bounding_box(self, prim)` (line 69): Compute the axis-aligned bounding box of a USD primitive subtree.
-  - `def save_usd_file_arrangement(self, new_stage_name, usd_file_names)` (line 129): Create a spatial grid arrangement of objects from multiple USD files.
-  - `def merge_usd_files(self, output_filename, input_filenames_list)` (line 252): Merge multiple USD files into a single comprehensive USD file.
-  - `def merge_usd_files_flattened(self, output_filename, input_filenames_list)` (line 443): Merge multiple USD files using references and flattening.
-  - `def list_mesh_primvars(self, stage_or_path, mesh_path, time_code=None)` (line 568): List all primvars on a USD mesh with metadata.
-  - `def pick_color_primvar(self, primvar_infos, keywords=('strain', 'stress'))` (line 663): Select a primvar for coloring based on keywords and preferences.
-  - `def apply_colormap_from_primvar(self, stage_or_path, mesh_path, source_primvar, *, cmap='viridis', time_codes=None, intensity_range=None, use_sigmoid_scale=False, write_default_at_t0=True, bind_vertex_color_material=True)` (line 717): Apply colormap visualization by converting a primvar to displayColor.
-  - `def set_solid_display_color(self, stage_or_path, mesh_path, color, *, time_codes=None, bind_vertex_color_material=True)` (line 1013): Set a constant (solid) displayColor for a mesh.
-  - `def list_mesh_paths_under(self, stage_or_path, parent_path='/World/Meshes')` (line 1106): List paths of all mesh prims under a parent path.
-  - `def repair_mesh_primvar_element_sizes(self, stage_or_path, mesh_path, *, time_code=None, save=True)` (line 1133): Repair missing/incorrect primvar elementSize metadata for a mesh.
+- **class USDTools** (line 28): Utilities for manipulating Universal Scene Description (USD) files.
+  - `def __init__(self, log_level=logging.INFO)` (line 64): Initialize the USDTools class.
+  - `def load_usd_as_vtk(self, usd_file, prim_path='/World', time_code=None)` (line 72): Load USD mesh geometry as a PyVista ``PolyData``.
+  - `def get_subtree_bounding_box(self, prim)` (line 218): Compute the axis-aligned bounding box of a USD primitive subtree.
+  - `def save_usd_file_arrangement(self, new_stage_name, usd_file_names)` (line 278): Create a spatial grid arrangement of objects from multiple USD files.
+  - `def merge_usd_files(self, output_filename, input_filenames_list)` (line 404): Merge multiple USD files into a single comprehensive USD file.
+  - `def merge_usd_files_flattened(self, output_filename, input_filenames_list)` (line 609): Merge multiple USD files using references and flattening.
+  - `def list_mesh_primvars(self, stage_or_path, mesh_path, time_code=None)` (line 737): List all primvars on a USD mesh with metadata.
+  - `def pick_color_primvar(self, primvar_infos, keywords=('strain', 'stress'))` (line 832): Select a primvar for coloring based on keywords and preferences.
+  - `def apply_colormap_from_primvar(self, stage_or_path, mesh_path, source_primvar, *, cmap='viridis', time_codes=None, intensity_range=None, use_sigmoid_scale=False, write_default_at_t0=True, bind_vertex_color_material=True)` (line 886): Apply colormap visualization by converting a primvar to displayColor.
+  - `def set_solid_display_color(self, stage_or_path, mesh_path, color, *, time_codes=None, bind_vertex_color_material=True)` (line 1182): Set a constant (solid) displayColor for a mesh.
+  - `def list_mesh_paths_under(self, stage_or_path, parent_path='/World/Meshes')` (line 1275): List paths of all mesh prims under a parent path.
+  - `def repair_mesh_primvar_element_sizes(self, stage_or_path, mesh_path, *, time_code=None, save=True)` (line 1302): Repair missing/incorrect primvar elementSize metadata for a mesh.
 
 ## src/physiomotion4d/vtk_to_usd/converter.py
 
-- `def convert_vtk_file(vtk_file, output_usd_file, *, data_basename=None, mesh_name='Mesh', extract_surface=True, settings=None, material=None)` (line 14): Convert one VTK file to one USD stage.
+- `def convert_vtk_file(vtk_file, output_usd_file, *, data_basename=None, mesh_name='Mesh', extract_surface=True, settings=None, material=None)` (line 15): Convert one VTK file to one USD stage.
 
 ## src/physiomotion4d/vtk_to_usd/data_structures.py
 
@@ -367,8 +375,8 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
   - `def __init__(self, stage, materials_scope_path='/World/Looks')` (line 23): Initialize material manager.
   - `def create_material(self, mat_data, time_code=None)` (line 37): Create a UsdPreviewSurface material.
   - `def bind_material(self, geom_prim, material)` (line 133): Bind a material to a geometry prim.
-  - `def get_or_create_material(self, mat_data, time_code=None)` (line 149): Get existing material from cache or create new one.
-  - `def create_default_material(self, name='default', color=(0.8, 0.8, 0.8))` (line 165): Create a simple default material.
+  - `def get_or_create_material(self, mat_data, time_code=None)` (line 152): Get existing material from cache or create new one.
+  - `def create_default_material(self, name='default', color=(0.8, 0.8, 0.8))` (line 168): Create a simple default material.
 
 ## src/physiomotion4d/vtk_to_usd/mesh_utils.py
 
@@ -376,12 +384,18 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 - `def split_mesh_data_by_cell_type(mesh_data, mesh_name)` (line 32): Split MeshData into one mesh per distinct face vertex count (cell type).
 - `def split_mesh_data_by_connectivity(mesh_data, mesh_name)` (line 280): Split MeshData into one mesh per connected component.
 
+## src/physiomotion4d/vtk_to_usd/primvar_derivations.py
+
+- `def compute_von_mises_stress(stress_tensor)` (line 49): Compute scalar von Mises stress from a row-major 9-component tensor.
+- `def derive_von_mises_from_stress(array)` (line 107): Derive a scalar VonMises primvar from any 9-component stress tensor.
+- `def derive_primvars(arrays)` (line 147): Apply every registered derivation to every array in ``arrays``.
+
 ## src/physiomotion4d/vtk_to_usd/usd_mesh_converter.py
 
 - **class UsdMeshConverter** (line 25): Converts MeshData to UsdGeomMesh with full feature support.
   - `def __init__(self, stage, settings, material_mgr)` (line 36): Initialize mesh converter.
-  - `def create_mesh(self, mesh_data, mesh_path, time_code=None, bind_material=True)` (line 53): Create a UsdGeomMesh from MeshData.
-  - `def create_time_varying_mesh(self, mesh_data_sequence, mesh_path, time_codes, bind_material=True)` (line 289): Create a mesh with time-varying attributes.
+  - `def create_mesh(self, mesh_data, mesh_path, time_code=None, bind_material=True)` (line 57): Create a UsdGeomMesh from MeshData.
+  - `def create_time_varying_mesh(self, mesh_data_sequence, mesh_path, time_codes, bind_material=True)` (line 329): Create a mesh with time-varying attributes.
 
 ## src/physiomotion4d/vtk_to_usd/usd_utils.py
 
@@ -392,8 +406,9 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 - `def get_sdf_value_type(data_type, num_components)` (line 173): Get appropriate SDF value type for primvar creation.
 - `def sanitize_primvar_name(name)` (line 220): Sanitize a name to be USD-compliant.
 - `def create_primvar(geom, array, array_name_prefix='', time_code=None)` (line 255): Create a USD primvar from a GenericArray.
-- `def triangulate_face(face_counts, face_indices)` (line 369): Triangulate polygonal faces.
-- `def compute_mesh_extent(points)` (line 409): Compute bounding box extent for a mesh.
+- `def triangulate_face(face_counts, face_indices)` (line 369): Triangulate polygonal faces using fan triangulation.
+- `def compute_mesh_extent(points)` (line 420): Compute bounding box extent for a mesh.
+- `def add_framing_camera(stage, *, parent_path='/World', name='Camera', bounds_min=None, bounds_max=None, focal_length_mm=50.0, horizontal_aperture_mm=36.0, distance_factor=3.0)` (line 432): Define a USD camera that frames stage geometry with tight clipping planes.
 
 ## src/physiomotion4d/vtk_to_usd/vtk_reader.py
 
@@ -411,17 +426,17 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 
 - **class WorkflowConvertCTToVTK** (line 58): Segment a CT image and produce per-anatomy-group VTK surfaces and meshes.
   - `def __init__(self, segmentation_method='total_segmentator', log_level=logging.INFO)` (line 98): Initialize the workflow.
-  - `def run_workflow(self, input_image, contrast_enhanced_study=False, anatomy_groups=None)` (line 240): Segment the CT image and extract per-anatomy-group VTK objects.
-  - `def save_surfaces(surfaces, output_dir, prefix='')` (line 343): Save each group surface to its own VTP file.
-  - `def save_meshes(meshes, output_dir, prefix='')` (line 370): Save each group voxel mesh to its own VTU file.
-  - `def save_combined_surface(surfaces, output_dir, prefix='')` (line 396): Merge all group surfaces into a single VTP file.
-  - `def save_combined_mesh(meshes, output_dir, prefix='')` (line 431): Merge all group meshes into a single VTU file.
+  - `def run_workflow(self, input_image, contrast_enhanced_study=False, anatomy_groups=None)` (line 241): Segment the CT image and extract per-anatomy-group VTK objects.
+  - `def save_surfaces(surfaces, output_dir, prefix='')` (line 344): Save each group surface to its own VTP file.
+  - `def save_meshes(meshes, output_dir, prefix='')` (line 371): Save each group voxel mesh to its own VTU file.
+  - `def save_combined_surface(surfaces, output_dir, prefix='')` (line 397): Merge all group surfaces into a single VTP file.
+  - `def save_combined_mesh(meshes, output_dir, prefix='')` (line 432): Merge all group meshes into a single VTU file.
 
 ## src/physiomotion4d/workflow_convert_heart_gated_ct_to_usd.py
 
 - **class WorkflowConvertHeartGatedCTToUSD** (line 28): Complete workflow for Heart-gated CT images to dynamic USD models.
-  - `def __init__(self, input_filenames, contrast_enhanced, output_directory, project_name, reference_image_filename=None, number_of_registration_iterations=1, registration_method='icon', log_level=logging.INFO)` (line 36): Initialize the Heart-gated CT to USD workflow.
-  - `def process(self)` (line 129): Execute the complete workflow from 4D CT to dynamic USD models.
+  - `def __init__(self, input_filenames, contrast_enhanced, output_directory, project_name, reference_image_filename=None, number_of_registration_iterations=1, registration_method='icon', log_level=logging.INFO, save_registered_images=True, save_registration_transforms=True, save_labelmaps=True)` (line 36): Initialize the Heart-gated CT to USD workflow.
+  - `def process(self)` (line 177): Execute the complete workflow from 4D CT to dynamic USD models.
 
 ## src/physiomotion4d/workflow_convert_vtk_to_usd.py
 
@@ -470,16 +485,16 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 
 ## tests/conftest.py
 
-- `def pytest_addoption(parser)` (line 36): Add custom command-line options for pytest.
-- `def pytest_configure(config)` (line 58): Configure pytest with custom markers and settings.
-- `def pytest_collection_modifyitems(config, items)` (line 85): Automatically skip experiment and tutorial tests unless their opt-in flags
-- `def pytest_runtest_logreport(report)` (line 110): Collect test timing information after each test completes.
-- `def pytest_terminal_summary(terminalreporter, exitstatus, config)` (line 135): Print comprehensive test timing report after all tests complete.
-- `def test_directories()` (line 297): Set up test directories for data and results.
-- `def download_test_data(test_directories)` (line 312): Download TruncalValve 4D CT data.
+- `def pytest_addoption(parser)` (line 35): Add custom command-line options for pytest.
+- `def pytest_configure(config)` (line 57): Configure pytest with custom markers and settings.
+- `def pytest_collection_modifyitems(config, items)` (line 84): Automatically skip experiment and tutorial tests unless their opt-in flags
+- `def pytest_runtest_logreport(report)` (line 109): Collect test timing information after each test completes.
+- `def pytest_terminal_summary(terminalreporter, exitstatus, config)` (line 134): Print comprehensive test timing report after all tests complete.
+- `def test_directories()` (line 296): Set up test directories for data and results.
+- `def download_test_data(test_directories)` (line 321): Download Slicer-Heart-CT data.
 - `def test_images(download_test_data, test_directories)` (line 348): Convert and resample 4D NRRD data; return pre-resampled time points.
-- `def test_labelmaps(segmenter_total_segmentator, test_images, test_directories)` (line 400): Segment each time point with TotalSegmentator and return result dicts.
-- `def test_transforms(registrar_ants, test_images, test_directories)` (line 441): Perform ANTs registration and return results.
+- `def test_labelmaps(segmenter_total_segmentator, test_images, test_directories)` (line 401): Segment each time point with TotalSegmentator and return result dicts.
+- `def test_transforms(registrar_ants, test_images, test_directories)` (line 442): Perform ANTs registration and return results.
 - `def segmenter_total_segmentator()` (line 497): Create a SegmentChestTotalSegmentator instance.
 - `def segmenter_simpleware()` (line 503): Create a SegmentHeartSimpleware instance.
 - `def contour_tools()` (line 509): Create a ContourTools instance.
@@ -488,6 +503,21 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 - `def registrar_icon()` (line 527): Create a RegisterImagesICON instance.
 - `def transform_tools()` (line 533): Create a TransformTools instance.
 
+## tests/test_anatomy_taxonomy.py
+
+- `def test_add_organ_creates_group_lazily()` (line 14)
+- `def test_group_names_preserve_insertion_order()` (line 23)
+- `def test_labels_in_group_unknown_returns_empty_dict()` (line 31)
+- `def test_labels_in_group_returns_copy_not_alias()` (line 37): Caller mutation of the returned dict must not corrupt the taxonomy.
+- `def test_all_labels_merges_every_group()` (line 46)
+- `def test_group_for_label_finds_by_organ_name()` (line 58)
+- `def test_group_for_label_unknown_falls_back_to_other()` (line 67)
+- `def test_group_for_id_finds_by_id()` (line 73)
+- `def test_fill_other_group_only_claims_unassigned_ids()` (line 82)
+- `def test_fill_other_group_idempotent_for_already_claimed_other()` (line 100): Calling fill_other_group twice must not duplicate or overwrite.
+- `def test_anatomy_group_dataclass_default_organs()` (line 110)
+- `def test_segment_anatomy_base_default_taxonomy_seeded()` (line 116): SegmentAnatomyBase seeds contrast (135) and soft_tissue (133).
+
 ## tests/test_cli_smoke.py
 
 - `def test_cli_help(module_name, monkeypatch, capsys)` (line 23): Each CLI module exits successfully for --help.
@@ -508,7 +538,7 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 ## tests/test_convert_nrrd_4d_to_3d.py
 
 - **class TestConvertNRRD4DTo3D** (line 17): Test suite for converting 4D NRRD to 3D time series.
-  - `def test_convert_4d_to_3d(self, download_test_data, test_directories)` (line 20): Test conversion of 4D NRRD to 3D time series (replicates notebook cell 3).
+  - `def test_convert_4d_to_3d(self, download_test_data, test_directories)` (line 20): Test conversion of 4D NRRD to 3D time series.
   - `def test_slice_files_created(self, download_test_data, test_directories)` (line 46): Test that all expected slice files are present after conversion.
   - `def test_load_nrrd_4d(self, download_test_data)` (line 67): Test loading 4D NRRD file.
   - `def test_save_3d_images(self, download_test_data, test_directories)` (line 80): Test saving 3D images from 4D NRRD.
@@ -530,16 +560,22 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 - **class TestSyntheticConversion** (line 425): Synthetic (no-disk-data) tests for ConvertVTKToUSD.
   - `def test_single_frame_prim_has_time_sample(self, tmp_path)` (line 438): Single-frame _convert_unified() must author one time sample, not a static prim.
   - `def test_static_merge_prim_names_use_data_basename(self, tmp_path)` (line 454): Static-merge prims must be named {data_basename}_{i}, not Mesh_{i}.
-  - `def test_mask_ids_basic_produces_per_label_prims(self, tmp_path)` (line 486): mask_ids must produce one USD prim per label; no unified /Mesh prim.
-  - `def test_structured_grid_extracts_surface(self, tmp_path)` (line 508): StructuredGrid input is surface-extracted when convert_to_surface is true.
-  - `def test_mask_ids_missing_label_filters_time_codes(self, tmp_path)` (line 524): Time codes for a label must be filtered to frames where it actually appears.
-  - `def test_mask_ids_missing_boundary_labels_falls_back(self, tmp_path)` (line 559): Mesh without boundary_labels array falls back to a 'default' prim.
+  - `def test_mask_ids_basic_produces_per_label_prims(self, tmp_path)` (line 486): mask_ids must produce one USD prim per label grouped under /Anatomy
+  - `def test_structured_grid_extracts_surface(self, tmp_path)` (line 512): StructuredGrid input is surface-extracted when convert_to_surface is true.
+  - `def test_mask_ids_missing_label_filters_time_codes(self, tmp_path)` (line 528): Time codes for a label must be filtered to frames where it actually appears.
+  - `def test_mask_ids_missing_boundary_labels_falls_back(self, tmp_path)` (line 563): Mesh without boundary_labels array falls back to a 'default' prim.
+  - `def test_mask_ids_groups_by_segmenter_type(self, tmp_path)` (line 578): When a segmenter is supplied, labels are grouped under their
 
 ## tests/test_download_heart_data.py
 
-- **class TestDownloadHeartData** (line 15): Test suite for downloading and converting Slicer-Heart-CT data.
-  - `def test_directories_created(self, test_directories)` (line 18): Test that directories are created successfully.
-  - `def test_data_downloaded(self, download_test_data, test_directories)` (line 28): Test that the TruncalValve 4D CT data file is downloaded.
+- **class TestDataDownloadTools** (line 16): Synthetic tests for dataset verification helpers.
+  - `def test_verify_slicer_heart_ct_data(self, tmp_path)` (line 19): Verify Slicer data by expected `TruncalValve_4DCT.seq.nrrd` filename.
+  - `def test_verify_kcl_heart_model_data(self, tmp_path)` (line 29): Verify KCL data by expected average mesh and input mesh filenames.
+  - `def test_verify_dirlab_4dct_data(self, tmp_path)` (line 41): Verify DirLab data by supported Case1 phase image layouts.
+  - `def test_verify_chop_valve_4d_data(self, tmp_path)` (line 52): Verify CHOP data by expected CT or valve time-series paths.
+- **class TestDownloadHeartData** (line 64): Test suite for downloading and converting Slicer-Heart-CT data.
+  - `def test_directories_created(self, test_directories)` (line 67): Test that directories are created successfully.
+  - `def test_data_downloaded(self, download_test_data, test_directories)` (line 91): Test that the TruncalValve 4D CT data file is downloaded.
 
 ## tests/test_experiments.py
 
@@ -663,22 +699,22 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 
 - **class TestSegmentChestTotalSegmentator** (line 21): Test suite for TotalSegmentator chest CT segmentation.
   - `def test_segmenter_initialization(self, segmenter_total_segmentator)` (line 24): Test that SegmentChestTotalSegmentator initializes correctly.
-  - `def test_segment_single_image(self, segmenter_total_segmentator, test_images, test_directories)` (line 61): Test segmentation on a single time point.
-  - `def test_segment_multiple_images(self, segmenter_total_segmentator, test_images, test_directories)` (line 119): Test segmentation on two time points.
-  - `def test_anatomy_group_masks(self, segmenter_total_segmentator, test_images)` (line 150): Test that anatomy group masks are created correctly.
-  - `def test_contrast_detection(self, segmenter_total_segmentator, test_images)` (line 194): Test contrast detection functionality.
-  - `def test_preprocessing(self, segmenter_total_segmentator, test_images)` (line 225): Test preprocessing functionality.
-  - `def test_postprocessing(self, segmenter_total_segmentator, test_images)` (line 249): Test postprocessing functionality.
+  - `def test_segment_single_image(self, segmenter_total_segmentator, test_images, test_directories)` (line 55): Test segmentation on a single time point.
+  - `def test_segment_multiple_images(self, segmenter_total_segmentator, test_images, test_directories)` (line 113): Test segmentation on two time points.
+  - `def test_anatomy_group_masks(self, segmenter_total_segmentator, test_images)` (line 144): Test that anatomy group masks are created correctly.
+  - `def test_contrast_detection(self, segmenter_total_segmentator, test_images)` (line 188): Test contrast detection functionality.
+  - `def test_preprocessing(self, segmenter_total_segmentator, test_images)` (line 219): Test preprocessing functionality.
+  - `def test_postprocessing(self, segmenter_total_segmentator, test_images)` (line 243): Test postprocessing functionality.
 
 ## tests/test_segment_heart_simpleware.py
 
 - **class TestSegmentHeartSimpleware** (line 32): Test suite for SegmentHeartSimpleware (Simpleware Medical ASCardio).
   - `def test_segmenter_initialization(self, segmenter_simpleware)` (line 35): Test that SegmentHeartSimpleware initializes correctly.
-  - `def test_set_simpleware_executable_path(self, segmenter_simpleware)` (line 61): Test setting custom Simpleware executable path.
-  - `def test_segment_single_image(self, segmenter_simpleware, test_images, test_directories)` (line 74): Test segmentation on a cardiac CT time point.
-  - `def test_anatomy_group_masks(self, segmenter_simpleware, test_images)` (line 129): Test that anatomy group masks are created (heart, vessels, etc.).
-  - `def test_contrast_detection(self, segmenter_simpleware, test_images)` (line 166): Test contrast mask is returned (base class behavior).
-  - `def test_postprocessing(self, segmenter_simpleware, test_images)` (line 182): Test that output labelmap matches input size and spacing.
+  - `def test_set_simpleware_executable_path(self, segmenter_simpleware)` (line 68): Test setting custom Simpleware executable path.
+  - `def test_segment_single_image(self, segmenter_simpleware, test_images, test_directories)` (line 81): Test segmentation on a cardiac CT time point.
+  - `def test_anatomy_group_masks(self, segmenter_simpleware, test_images)` (line 145): Test that anatomy group masks are created (heart, vessels, etc.).
+  - `def test_contrast_detection(self, segmenter_simpleware, test_images)` (line 182): Test contrast mask is returned (base class behavior).
+  - `def test_postprocessing(self, segmenter_simpleware, test_images)` (line 198): Test that output labelmap matches input size and spacing.
 
 ## tests/test_transform_tools.py
 
@@ -705,24 +741,18 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 
 ## tests/test_tutorials.py
 
-- `def test_testtools_results_output_dir_override(tmp_path)` (line 78): Store result artifacts in an explicit directory when requested.
-- `def test_tutorial_01_contour_png_mesh_uses_current_run_results()` (line 103): Select current in-memory contours instead of disk VTP outputs.
-- `def test_tutorial_01_reference_png_uses_workflow_fixed_image(tmp_path)` (line 119): Select the actual workflow reference image over cached slice images.
-- `def test_tutorial_01_overlay_uses_workflow_fixed_segmentation(tmp_path)` (line 134): Select the current fixed labelmap over cached slice labelmaps.
-- `def test_tutorial_01_overlay_falls_back_to_fixed_image_mask(tmp_path)` (line 149): Read fixed_image_mask.mha before stale slice labelmap files.
-- **class TestTutorial01HeartGatedCTToUSD** (line 176): End-to-end test for tutorial_01_heart_gated_ct_to_usd.py.
-  - `def test_run(self, test_directories)` (line 181)
-- **class TestTutorial02CTToVTK** (line 210): End-to-end test for tutorial_02_ct_to_vtk.py.
-  - `def test_run(self, test_directories)` (line 215)
-- **class TestTutorial03CreateStatisticalModel** (line 243): End-to-end test for tutorial_03_create_statistical_model.py.
-  - `def test_run(self, test_directories)` (line 248)
-- `def test_tutorial_04_extract_surface_uses_dataset_surface()` (line 281): Use the robust dataset_surface algorithm for VTK surface extraction.
-- **class TestTutorial04FitStatisticalModelToPatient** (line 296): End-to-end test for tutorial_04_fit_statistical_model_to_patient.py.
-  - `def test_run(self, test_directories)` (line 301)
-- **class TestTutorial05VTKToUSD** (line 356): End-to-end test for tutorial_05_vtk_to_usd.py.
-  - `def test_run(self, test_directories)` (line 361)
-- **class TestTutorial06ReconstructHighres4DCT** (line 406): End-to-end test for tutorial_06_reconstruct_highres_4d_ct.py.
-  - `def test_run(self, test_directories)` (line 411)
+- **class TestTutorial01HeartGatedCTToUSD** (line 79): End-to-end test for tutorial_01_heart_gated_ct_to_usd.py.
+  - `def test_run(self, test_directories)` (line 84)
+- **class TestTutorial02CTToVTK** (line 107): End-to-end test for tutorial_02_ct_to_vtk.py.
+  - `def test_run(self, test_directories)` (line 112)
+- **class TestTutorial03CreateStatisticalModel** (line 134): End-to-end test for tutorial_03_create_statistical_model.py.
+  - `def test_run(self, test_directories)` (line 139)
+- **class TestTutorial04FitStatisticalModelToPatient** (line 162): End-to-end test for tutorial_04_fit_statistical_model_to_patient.py.
+  - `def test_run(self, test_directories)` (line 167)
+- **class TestTutorial05VTKToUSD** (line 206): End-to-end test for tutorial_05_vtk_to_usd.py.
+  - `def test_run(self, test_directories)` (line 211)
+- **class TestTutorial06ReconstructHighres4DCT** (line 247): End-to-end test for tutorial_06_reconstruct_highres_4d_ct.py.
+  - `def test_run(self, test_directories)` (line 252)
 
 ## tests/test_usd_merge.py
 
@@ -752,31 +782,32 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 
 ## tests/test_vtk_to_usd_library.py
 
-- `def get_data_dir()` (line 24): Get the data directory path.
-- `def check_kcl_heart_data()` (line 31): Check if KCL Heart Model data is available.
-- `def get_or_create_average_surface(test_directories)` (line 38): Get or create average_surface.vtp from average_mesh.vtk.
-- `def kcl_average_surface(test_directories)` (line 65): Fixture providing the KCL average heart surface.
-- **class TestFromFilesValidation** (line 78): Synthetic tests for ConvertVTKToUSD.from_files().
-  - `def test_time_codes_length_mismatch_raises(self, tmp_path)` (line 81): from_files() must reject time_codes whose length != len(vtk_files).
-  - `def test_time_codes_non_monotone_raises(self, tmp_path)` (line 91): from_files() must reject time_codes that decrease between frames.
-  - `def test_time_codes_equal_consecutive_is_valid(self, tmp_path)` (line 101): Equal consecutive time codes are non-decreasing and must not raise.
-  - `def test_from_files_single_file_writes_static_mesh(self, tmp_path)` (line 113): A single-file converter writes a static mesh with no time range.
-  - `def test_from_files_static_merge_writes_separate_meshes(self, tmp_path)` (line 125): static_merge=True treats files as static objects, not time samples.
-- **class TestSyntheticConversion** (line 144): Synthetic ConvertVTKToUSD tests that do not require downloaded data.
-  - `def test_inspect_file_reports_public_summary(self, tmp_path)` (line 147): inspect_file() reports geometry, bounds, arrays, and cell types.
-  - `def test_inspect_file_reports_empty_mesh(self, tmp_path)` (line 167): inspect_file() reports empty meshes without raising.
-  - `def test_file_primvar_preservation(self, tmp_path)` (line 183): Point arrays in a VTP file are preserved as USD primvars.
-  - `def test_time_series_conversion(self, tmp_path)` (line 202): Multiple VTP files write point time samples and stage time metadata.
-- **class TestVTKToUSDConversion** (line 226): Test ConvertVTKToUSD on optional real VTK data.
-  - `def test_single_file_conversion(self, test_directories, kcl_average_surface)` (line 229): Test converting a single VTK file to USD.
-  - `def test_conversion_with_material(self, test_directories, kcl_average_surface)` (line 248): Test conversion with a custom solid color material.
-  - `def test_conversion_settings(self, test_directories, kcl_average_surface)` (line 276): Test that ConvertVTKToUSD applies correct default stage metadata.
-- **class TestIntegration** (line 293): Integration tests combining multiple features.
-  - `def test_end_to_end_conversion(self, test_directories, kcl_average_surface)` (line 296): Test complete conversion workflow with all features.
-- **class TestUnitScaling** (line 319): Verify that VTK mm coordinates are converted to USD meter coordinates.
-  - `def test_mm_to_m_point_scaling(self, tmp_path)` (line 322): Points written to USD must be 0.001x their original mm values.
-  - `def test_normals_remain_unit_length(self, tmp_path)` (line 342): Normal vectors must not be scaled.
-  - `def test_stage_meters_per_unit(self, tmp_path)` (line 362): Stage metersPerUnit metadata must be 1.0.
+- `def get_data_dir()` (line 27): Get the data directory path.
+- `def check_kcl_heart_data()` (line 34): Check if KCL Heart Model data is available.
+- `def get_or_create_average_surface(test_directories)` (line 41): Get or create average_surface.vtp from average_mesh.vtk.
+- `def kcl_average_surface(test_directories)` (line 68): Fixture providing the KCL average heart surface.
+- **class TestFromFilesValidation** (line 81): Synthetic tests for ConvertVTKToUSD.from_files().
+  - `def test_time_codes_length_mismatch_raises(self, tmp_path)` (line 84): from_files() must reject time_codes whose length != len(vtk_files).
+  - `def test_time_codes_non_monotone_raises(self, tmp_path)` (line 94): from_files() must reject time_codes that decrease between frames.
+  - `def test_time_codes_equal_consecutive_is_valid(self, tmp_path)` (line 104): Equal consecutive time codes are non-decreasing and must not raise.
+  - `def test_from_files_single_file_writes_static_mesh(self, tmp_path)` (line 116): A single-file converter writes a static mesh with no time range.
+  - `def test_openusd_screenshot_uses_vtk_loader(self, tmp_path)` (line 128): Render a tiny OpenUSD mesh through TestTools without USD imaging plugins.
+  - `def test_from_files_static_merge_writes_separate_meshes(self, tmp_path)` (line 168): static_merge=True treats files as static objects, not time samples.
+- **class TestSyntheticConversion** (line 187): Synthetic ConvertVTKToUSD tests that do not require downloaded data.
+  - `def test_inspect_file_reports_public_summary(self, tmp_path)` (line 190): inspect_file() reports geometry, bounds, arrays, and cell types.
+  - `def test_inspect_file_reports_empty_mesh(self, tmp_path)` (line 210): inspect_file() reports empty meshes without raising.
+  - `def test_file_primvar_preservation(self, tmp_path)` (line 226): Point arrays in a VTP file are preserved as USD primvars.
+  - `def test_time_series_conversion(self, tmp_path)` (line 245): Multiple VTP files write point time samples and stage time metadata.
+- **class TestVTKToUSDConversion** (line 269): Test ConvertVTKToUSD on optional real VTK data.
+  - `def test_single_file_conversion(self, test_directories, kcl_average_surface)` (line 272): Test converting a single VTK file to USD.
+  - `def test_conversion_with_material(self, test_directories, kcl_average_surface)` (line 291): Test conversion with a custom solid color material.
+  - `def test_conversion_settings(self, test_directories, kcl_average_surface)` (line 319): Test that ConvertVTKToUSD applies correct default stage metadata.
+- **class TestIntegration** (line 336): Integration tests combining multiple features.
+  - `def test_end_to_end_conversion(self, test_directories, kcl_average_surface)` (line 339): Test complete conversion workflow with all features.
+- **class TestUnitScaling** (line 362): Verify that VTK mm coordinates are converted to USD meter coordinates.
+  - `def test_mm_to_m_point_scaling(self, tmp_path)` (line 365): Points written to USD must be 0.001x their original mm values.
+  - `def test_normals_remain_unit_length(self, tmp_path)` (line 385): Normal vectors must not be scaled.
+  - `def test_stage_meters_per_unit(self, tmp_path)` (line 405): Stage metersPerUnit metadata must be 1.0.
 
 ## tests/test_workflow_fit_statistical_model_to_patient.py
 
@@ -784,30 +815,6 @@ _Re-run `py utils/generate_api_map.py` whenever public APIs change._
 - `def test_transform_model_applies_staged_transform()` (line 61): Transform helper updates mesh points with image shape (Z, Y, X) = (3, 3, 3).
 - `def test_transform_model_preserves_unstructured_grid_topology()` (line 93): Transform helper preserves cells with image shape (Z, Y, X) = (3, 3, 3).
 
-## tutorials/tutorial_01_heart_gated_ct_to_usd.py
-
-- `def run_tutorial(data_dir, output_dir, *, registration_method='ants', log_level=logging.INFO)` (line 176): Run Tutorial 1: Heart-Gated CT to Animated USD.
-
-## tutorials/tutorial_02_ct_to_vtk.py
-
-- `def run_tutorial(data_dir, output_dir, *, log_level=logging.INFO)` (line 122): Run Tutorial 2: CT Segmentation to VTK Surfaces.
-
-## tutorials/tutorial_03_create_statistical_model.py
-
-- `def run_tutorial(data_dir, output_dir, *, pca_components=10, max_samples=20, log_level=logging.INFO)` (line 93): Run Tutorial 3: Create a PCA Statistical Shape Model.
-
-## tutorials/tutorial_04_fit_statistical_model_to_patient.py
-
-- `def run_tutorial(data_dir, output_dir, *, pca_json=None, log_level=logging.INFO)` (line 101): Run Tutorial 4: Fit Statistical Shape Model to Patient Data.
-
-## tutorials/tutorial_05_vtk_to_usd.py
-
-- `def run_tutorial(data_dir, output_dir, *, vtk_file=None, log_level=logging.INFO)` (line 80): Run Tutorial 5: VTK Surface Series to Animated USD.
-
-## tutorials/tutorial_06_reconstruct_highres_4d_ct.py
-
-- `def run_tutorial(data_dir, output_dir, *, case=1, max_frames=4, registration_method='ants', log_level=logging.INFO)` (line 97): Run Tutorial 6: Reconstruct High-Resolution 4D CT.
-
 ## utils/ai_agent_github_reviews.py
 
 - `def git_fetch(repo_root, remote, branch)` (line 71): Run ``git fetch <remote> <branch>``, printing progress.
diff --git a/docs/README.md b/docs/README.md
index 0c7fbb4..3c09ca6 100644
--- a/docs/README.md
+++ b/docs/README.md
@@ -2,15 +2,15 @@
 
 This directory contains the source files for PhysioMotion4D's documentation, which is built using [Sphinx](https://www.sphinx-doc.org/) and hosted on [ReadTheDocs](https://readthedocs.org/).
 
-## 🎉 Recently Updated!
+## Recently Updated!
 
 The documentation has been restructured with:
-- ✅ Modern table-of-contents sidebar with search
-- ✅ Separate pages for each module (33 new API files)
-- ✅ Navigation widgets (prev/next/up) on all pages
-- ✅ Comprehensive API reference organized by functionality
-- ✅ Enhanced search capabilities
-- ✅ Modern custom CSS styling
+- Modern table-of-contents sidebar with search
+- Separate pages for each module (33 new API files)
+- Navigation widgets (prev/next/up) on all pages
+- Comprehensive API reference organized by functionality
+- Enhanced search capabilities
+- Modern custom CSS styling
 
 ## Building Documentation Locally
 
@@ -66,7 +66,7 @@ docs/
 ├── quickstart.rst               # Quick start guide
 ├── examples.rst                 # Code examples
 │
-├── api/                         # 📚 API Reference (NEW STRUCTURE!)
+├── api/                         # API Reference (NEW STRUCTURE!)
 │   ├── index.rst               # Main API hub
 │   ├── base.rst                # Core base class
 │   ├── workflows.rst           # Workflow classes
@@ -84,13 +84,13 @@ docs/
 │   ├── usd/                    # USD generation (6 files)
 │   └── utilities/              # Utilities (5 files)
 │
-├── developer/                   # 👨‍💻 Developer guides
+├── developer/                   # ‍Developer guides
 │   ├── architecture.rst
 │   ├── extending.rst
 │   ├── workflows.rst
 │   └── core.rst
 │
-├── cli_scripts/                 # 🔧 CLI documentation
+├── cli_scripts/                 # CLI documentation
 ├── contributing.rst
 ├── testing.rst
 └── ...
diff --git a/docs/api/base.rst b/docs/api/base.rst
index 7d40a4d..2df904a 100644
--- a/docs/api/base.rst
+++ b/docs/api/base.rst
@@ -2,6 +2,7 @@
 Base Class
 ====================================
 
+.. module:: physiomotion4d.physiomotion4d_base
 .. currentmodule:: physiomotion4d
 
 ``PhysioMotion4DBase`` provides the shared logging behavior used by workflow,
diff --git a/docs/api/cli/convert_ct_to_vtk.rst b/docs/api/cli/convert_ct_to_vtk.rst
new file mode 100644
index 0000000..e910d8d
--- /dev/null
+++ b/docs/api/cli/convert_ct_to_vtk.rst
@@ -0,0 +1,10 @@
+=======================
+convert_ct_to_vtk (CLI)
+=======================
+
+.. automodule:: physiomotion4d.cli.convert_ct_to_vtk
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+See :doc:`../../cli_scripts/overview` for usage.
diff --git a/docs/api/cli/convert_heart_gated_ct_to_usd.rst b/docs/api/cli/convert_heart_gated_ct_to_usd.rst
new file mode 100644
index 0000000..2f1af1f
--- /dev/null
+++ b/docs/api/cli/convert_heart_gated_ct_to_usd.rst
@@ -0,0 +1,10 @@
+===================================
+convert_heart_gated_ct_to_usd (CLI)
+===================================
+
+.. automodule:: physiomotion4d.cli.convert_heart_gated_ct_to_usd
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+See :doc:`../../cli_scripts/heart_gated_ct` for usage examples.
diff --git a/docs/api/cli/convert_vtk_to_usd.rst b/docs/api/cli/convert_vtk_to_usd.rst
new file mode 100644
index 0000000..cd663da
--- /dev/null
+++ b/docs/api/cli/convert_vtk_to_usd.rst
@@ -0,0 +1,10 @@
+========================
+convert_vtk_to_usd (CLI)
+========================
+
+.. automodule:: physiomotion4d.cli.convert_vtk_to_usd
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+See :doc:`../../cli_scripts/vtk_to_usd` for usage.
diff --git a/docs/api/cli/create_statistical_model.rst b/docs/api/cli/create_statistical_model.rst
new file mode 100644
index 0000000..f56efce
--- /dev/null
+++ b/docs/api/cli/create_statistical_model.rst
@@ -0,0 +1,10 @@
+==============================
+create_statistical_model (CLI)
+==============================
+
+.. automodule:: physiomotion4d.cli.create_statistical_model
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+See :doc:`../../cli_scripts/create_statistical_model` for usage.
diff --git a/docs/api/cli/fit_statistical_model_to_patient.rst b/docs/api/cli/fit_statistical_model_to_patient.rst
new file mode 100644
index 0000000..16a9338
--- /dev/null
+++ b/docs/api/cli/fit_statistical_model_to_patient.rst
@@ -0,0 +1,10 @@
+======================================
+fit_statistical_model_to_patient (CLI)
+======================================
+
+.. automodule:: physiomotion4d.cli.fit_statistical_model_to_patient
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+See :doc:`../../cli_scripts/fit_statistical_model_to_patient` for usage.
diff --git a/docs/api/cli/index.rst b/docs/api/cli/index.rst
new file mode 100644
index 0000000..61799e0
--- /dev/null
+++ b/docs/api/cli/index.rst
@@ -0,0 +1,33 @@
+====================
+CLI Entry-Point API
+====================
+
+.. module:: physiomotion4d.cli
+.. currentmodule:: physiomotion4d.cli
+
+The ``physiomotion4d.cli`` subpackage contains the entry-point scripts that
+back the installed ``physiomotion4d-*`` console commands. Each module exposes
+a ``main()`` function that parses ``argparse`` arguments and dispatches into
+the corresponding workflow class.
+
+User-facing documentation for the command-line tools (flags, examples, recipes)
+lives under :doc:`../../cli_scripts/overview`. This section documents the
+Python entry-point modules themselves so they are reachable from the Python
+Module Index.
+
+.. toctree::
+   :maxdepth: 1
+
+   convert_heart_gated_ct_to_usd
+   convert_ct_to_vtk
+   convert_vtk_to_usd
+   create_statistical_model
+   fit_statistical_model_to_patient
+   reconstruct_highres_4d_ct
+   visualize_pca_modes
+
+See Also
+========
+
+* :doc:`../../cli_scripts/overview`
+* :doc:`../workflows`
diff --git a/docs/api/cli/reconstruct_highres_4d_ct.rst b/docs/api/cli/reconstruct_highres_4d_ct.rst
new file mode 100644
index 0000000..7072fd4
--- /dev/null
+++ b/docs/api/cli/reconstruct_highres_4d_ct.rst
@@ -0,0 +1,10 @@
+===============================
+reconstruct_highres_4d_ct (CLI)
+===============================
+
+.. automodule:: physiomotion4d.cli.reconstruct_highres_4d_ct
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+See :doc:`../../cli_scripts/4dct_reconstruction` for usage.
diff --git a/docs/api/cli/visualize_pca_modes.rst b/docs/api/cli/visualize_pca_modes.rst
new file mode 100644
index 0000000..7bf270a
--- /dev/null
+++ b/docs/api/cli/visualize_pca_modes.rst
@@ -0,0 +1,10 @@
+=========================
+visualize_pca_modes (CLI)
+=========================
+
+.. automodule:: physiomotion4d.cli.visualize_pca_modes
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+See :doc:`../../cli_scripts/create_statistical_model` for usage in context.
diff --git a/docs/api/index.rst b/docs/api/index.rst
index 0d2b044..42bd8ab 100644
--- a/docs/api/index.rst
+++ b/docs/api/index.rst
@@ -17,6 +17,7 @@ This section provides detailed documentation for all PhysioMotion4D classes, fun
    model_registration/index
    usd/index
    utilities/index
+   cli/index
 
 Quick Navigation
 ================
diff --git a/docs/api/model_registration/distance_maps.rst b/docs/api/model_registration/distance_maps.rst
index cb36a34..efdfdd8 100644
--- a/docs/api/model_registration/distance_maps.rst
+++ b/docs/api/model_registration/distance_maps.rst
@@ -2,6 +2,7 @@
 Distance Map Registration
 ====================================
 
+.. module:: physiomotion4d.register_models_distance_maps
 .. currentmodule:: physiomotion4d
 
 Register models using distance field optimization.
diff --git a/docs/api/model_registration/icp.rst b/docs/api/model_registration/icp.rst
index 71f6ba9..b10ecf1 100644
--- a/docs/api/model_registration/icp.rst
+++ b/docs/api/model_registration/icp.rst
@@ -2,6 +2,7 @@
 ICP (Iterative Closest Point)
 ====================================
 
+.. module:: physiomotion4d.register_models_icp
 .. currentmodule:: physiomotion4d
 
 Pure Python implementation of Iterative Closest Point registration.
diff --git a/docs/api/model_registration/icp_itk.rst b/docs/api/model_registration/icp_itk.rst
index 593661d..39bc20e 100644
--- a/docs/api/model_registration/icp_itk.rst
+++ b/docs/api/model_registration/icp_itk.rst
@@ -2,6 +2,7 @@
 ICP with ITK Backend
 ====================================
 
+.. module:: physiomotion4d.register_models_icp_itk
 .. currentmodule:: physiomotion4d
 
 ICP registration using ITK's optimized implementation.
diff --git a/docs/api/model_registration/pca.rst b/docs/api/model_registration/pca.rst
index e345b8a..c8512a7 100644
--- a/docs/api/model_registration/pca.rst
+++ b/docs/api/model_registration/pca.rst
@@ -2,6 +2,7 @@
 PCA-Based Registration
 ====================================
 
+.. module:: physiomotion4d.register_models_pca
 .. currentmodule:: physiomotion4d
 
 Shape-based registration using principal component analysis.
diff --git a/docs/api/registration/ants.rst b/docs/api/registration/ants.rst
index ac4c075..c7deda9 100644
--- a/docs/api/registration/ants.rst
+++ b/docs/api/registration/ants.rst
@@ -2,6 +2,7 @@
 ANTs Registration
 =================
 
+.. module:: physiomotion4d.register_images_ants
 .. currentmodule:: physiomotion4d
 
 ``RegisterImagesANTs`` provides optimization-based deformable image
diff --git a/docs/api/registration/base.rst b/docs/api/registration/base.rst
index 3118c1b..055cbbb 100644
--- a/docs/api/registration/base.rst
+++ b/docs/api/registration/base.rst
@@ -2,6 +2,7 @@
 Registration Base Class
 ====================================
 
+.. module:: physiomotion4d.register_images_base
 .. currentmodule:: physiomotion4d
 
 Abstract base class for all image registration methods.
diff --git a/docs/api/registration/greedy.rst b/docs/api/registration/greedy.rst
new file mode 100644
index 0000000..3dcb989
--- /dev/null
+++ b/docs/api/registration/greedy.rst
@@ -0,0 +1,56 @@
+===================
+Greedy Registration
+===================
+
+.. module:: physiomotion4d.register_images_greedy
+.. currentmodule:: physiomotion4d
+
+``RegisterImagesGreedy`` provides fast CPU-based deformable image registration
+using the PICSL Greedy backend.
+
+Class Reference
+===============
+
+.. autoclass:: RegisterImagesGreedy
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Basic Registration
+==================
+
+.. code-block:: python
+
+   import itk
+
+   from physiomotion4d import RegisterImagesGreedy
+
+   fixed = itk.imread("reference.mha")
+   moving = itk.imread("moving.mha")
+
+   registrar = RegisterImagesGreedy()
+   registrar.set_modality("ct")
+   registrar.set_fixed_image(fixed)
+
+   result = registrar.register(moving)
+
+   forward_transform = result["forward_transform"]
+   inverse_transform = result["inverse_transform"]
+   registered = registrar.get_registered_image()
+
+Installation Note
+=================
+
+``RegisterImagesGreedy`` requires the ``picsl-greedy`` package, which is not
+installed by default:
+
+.. code-block:: bash
+
+   pip install picsl-greedy
+
+See Also
+========
+
+* :doc:`ants`
+* :doc:`icon`
+* :doc:`time_series`
diff --git a/docs/api/registration/icon.rst b/docs/api/registration/icon.rst
index f2df2d2..38bd4ba 100644
--- a/docs/api/registration/icon.rst
+++ b/docs/api/registration/icon.rst
@@ -2,6 +2,7 @@
 ICON Image Registration
 ================================
 
+.. module:: physiomotion4d.register_images_icon
 .. currentmodule:: physiomotion4d
 
 ``RegisterImagesICON`` performs deformable image registration using the
diff --git a/docs/api/registration/index.rst b/docs/api/registration/index.rst
index ed9e42b..8a472b0 100644
--- a/docs/api/registration/index.rst
+++ b/docs/api/registration/index.rst
@@ -12,6 +12,7 @@ PhysioMotion4D image registration classes align moving 3D images to a fixed
 
    base
    ants
+   greedy
    icon
    time_series
 
diff --git a/docs/api/registration/time_series.rst b/docs/api/registration/time_series.rst
index a23a105..20ccfdd 100644
--- a/docs/api/registration/time_series.rst
+++ b/docs/api/registration/time_series.rst
@@ -2,6 +2,7 @@
 Time-Series Registration
 ========================
 
+.. module:: physiomotion4d.register_time_series_images
 .. currentmodule:: physiomotion4d
 
 ``RegisterTimeSeriesImages`` registers ordered 3D image phases to a reference
diff --git a/docs/api/segmentation/base.rst b/docs/api/segmentation/base.rst
index 3aa1088..38e2eff 100644
--- a/docs/api/segmentation/base.rst
+++ b/docs/api/segmentation/base.rst
@@ -2,10 +2,13 @@
 Segmentation Base Class
 ========================
 
+.. module:: physiomotion4d.segment_anatomy_base
 .. currentmodule:: physiomotion4d
 
 ``SegmentAnatomyBase`` defines the shared chest-anatomy segmentation contract
-used by PhysioMotion4D segmentation implementations.
+used by PhysioMotion4D segmentation implementations. It owns an
+:class:`AnatomyTaxonomy` instance that subclasses populate to declare which
+anatomy groups (and which organ labels within each group) they produce.
 
 Class Reference
 ===============
@@ -30,23 +33,79 @@ Concrete segmenters accept an ITK image and return a dictionary of ITK images:
    segmenter = SegmentChestTotalSegmentator()
    masks = segmenter.segment(image, contrast_enhanced_study=True)
 
-   heart = masks["heart"]
    labelmap = masks["labelmap"]
+   if "heart" in masks:
+       heart = masks["heart"]
+
+The returned dictionary always contains ``"labelmap"`` plus one entry per
+anatomy group the segmenter registered in its taxonomy (and ``"other"`` for
+unclassified labels). **The exact key set is segmenter-specific** — callers
+must check membership (``"lung" in masks``) rather than assume a fixed
+schema. For example, :class:`SegmentChestTotalSegmentator` returns the full
+``heart, lung, bone, major_vessels, soft_tissue, contrast, other`` set,
+while :class:`SegmentHeartSimpleware` returns only the groups its
+ASCardio module actually populates (``heart``, ``major_vessels``,
+``soft_tissue``, ``contrast``, ``other``).
+
+Anatomy Taxonomy
+================
+
+The group-to-organ mapping is held by :class:`AnatomyTaxonomy`, a small
+data class shared between the segmenter and downstream renderers
+(:class:`USDAnatomyTools`, :class:`ConvertVTKToUSD`). It is independent of
+ITK and OpenUSD so segmentation code can be reasoned about without pulling
+in the rendering stack.
+
+.. autoclass:: AnatomyGroup
+   :members:
+
+.. autoclass:: AnatomyTaxonomy
+   :members:
+
+Typical usage from a subclass ``__init__``:
+
+.. code-block:: python
 
-Returned keys include ``labelmap``, ``lung``, ``heart``, ``major_vessels``,
-``bone``, ``soft_tissue``, ``other``, and ``contrast``.
+   class SegmentMySite(SegmentAnatomyBase):
+       def __init__(self):
+           super().__init__()
+           self.taxonomy.add_organ("heart", 51, "myocardium")
+           self.taxonomy.add_organ("heart", 61, "atrial_appendage_left")
+           self.taxonomy.add_organ("lung", 10, "lung_upper_lobe_left")
+           # ...
+           self._finalize_other_group()
+
+Downstream callers can introspect what a segmenter produces without
+running it:
+
+.. code-block:: python
+
+   tax = segmenter.taxonomy
+   print(tax.group_names())                       # ['heart', 'lung', ...]
+   print(tax.labels_in_group("heart"))            # {51: 'myocardium', ...}
+   print(tax.group_for_label("myocardium"))       # 'heart'
+   print(tax.all_labels())                        # full id -> name dict
 
 Extending Segmentation
 ======================
 
-New runtime segmentation classes should inherit from ``SegmentAnatomyBase`` or
-``PhysioMotion4DBase``, use ``log_info()`` / ``log_debug()``, and document the
-returned mask keys. Keep synthetic tests small and mark real-data tests with
-``requires_data``.
+New runtime segmentation classes should:
+
+1. Inherit from :class:`SegmentAnatomyBase` (or another :class:`PhysioMotion4DBase`
+   subclass if no anatomy taxonomy is needed).
+2. Populate ``self.taxonomy`` with ``add_organ`` calls in ``__init__``.
+3. Call ``self._finalize_other_group()`` once all groups have been registered.
+4. Use ``log_info()`` / ``log_debug()`` instead of ``print``.
+5. Document the key set the segmenter produces; downstream callers should
+   check membership rather than assume a fixed schema.
+
+Keep synthetic tests small and mark real-data tests with ``requires_data``.
 
 See Also
 ========
 
 * :doc:`totalsegmentator`
+* :doc:`simpleware`
 * :doc:`index`
 * :doc:`../../developer/segmentation`
+* :doc:`../usd/anatomy_tools` for the renderer side of the taxonomy
diff --git a/docs/api/segmentation/index.rst b/docs/api/segmentation/index.rst
index b621c81..bde3c2c 100644
--- a/docs/api/segmentation/index.rst
+++ b/docs/api/segmentation/index.rst
@@ -22,6 +22,7 @@ Quick Links
 **Segmentation Classes**:
    * :doc:`base` - Base class for all segmentation methods
    * :doc:`totalsegmentator` - TotalSegmentator implementation
+   * :doc:`simpleware` - Simpleware ASCardio cardiac segmentation
 
 Choosing a Method
 =================
@@ -56,6 +57,7 @@ Module Documentation
 
    base
    totalsegmentator
+   simpleware
 
 Common Operations
 =================
@@ -63,14 +65,16 @@ Common Operations
 Structure Extraction
 --------------------
 
-Extract individual anatomical structures from segmentation results:
+Extract individual anatomical structures from segmentation results. The key
+set returned by ``segment()`` is segmenter-specific (see :doc:`base` for the
+anatomy taxonomy contract), so check membership before accessing:
 
 .. code-block:: python
 
    result = segmenter.segment(ct_image)
-   heart_mask = result['heart']
-   lung_mask  = result['lung']
-   bone_mask  = result['bone']
+   for group in ("heart", "lung", "bone"):
+       if group in result:
+           itk.imwrite(result[group], f"{group}_mask.mha")
 
 Batch Processing
 ----------------
@@ -110,4 +114,4 @@ See Also
 
 .. rubric:: Navigation
 
-:doc:`../index` | :doc:`base` | :doc:`totalsegmentator`
+:doc:`../index` | :doc:`base` | :doc:`totalsegmentator` | :doc:`simpleware`
diff --git a/docs/api/segmentation/simpleware.rst b/docs/api/segmentation/simpleware.rst
new file mode 100644
index 0000000..887258a
--- /dev/null
+++ b/docs/api/segmentation/simpleware.rst
@@ -0,0 +1,68 @@
+==========================
+Simpleware Heart Segmenter
+==========================
+
+.. module:: physiomotion4d.segment_heart_simpleware
+.. currentmodule:: physiomotion4d
+
+``SegmentHeartSimpleware`` runs Synopsys Simpleware Medical's ASCardio module
+as an external process and returns the resulting heart and major-vessel masks
+as ITK images.
+
+Class Reference
+===============
+
+.. autoclass:: SegmentHeartSimpleware
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Basic Usage
+===========
+
+.. code-block:: python
+
+   import itk
+
+   from physiomotion4d import SegmentHeartSimpleware
+
+   image = itk.imread("chest_ct.nrrd")
+   segmenter = SegmentHeartSimpleware()
+   masks = segmenter.segment(image, contrast_enhanced_study=True)
+
+   heart = masks["heart"]
+   vessels = masks["major_vessels"]
+
+Returned Keys
+=============
+
+The ASCardio module segments cardiac anatomy only, so this segmenter's
+taxonomy registers a subset of the groups produced by
+:class:`SegmentChestTotalSegmentator`. The returned dictionary contains:
+
+* ``labelmap``
+* ``heart``
+* ``major_vessels``
+* ``soft_tissue`` (base-class placeholder for label id 133)
+* ``contrast`` (base-class placeholder for label id 135)
+* ``other`` (all unclaimed label ids in [1, 256))
+
+Keys such as ``lung`` and ``bone`` are **not** present. Callers that need
+those groups must either use a different segmenter or check membership
+(``"lung" in masks``) and handle the absence explicitly. See
+:doc:`base` for the full taxonomy contract.
+
+Installation Note
+=================
+
+``SegmentHeartSimpleware`` requires a licensed installation of Synopsys
+Simpleware Medical and its ASCardio module. The class invokes the Simpleware
+executable as a subprocess; the executable and helper script paths must be
+configured on the host system.
+
+See Also
+========
+
+* :doc:`base`
+* :doc:`totalsegmentator`
+* :doc:`index`
diff --git a/docs/api/segmentation/totalsegmentator.rst b/docs/api/segmentation/totalsegmentator.rst
index cf4c067..2b11776 100644
--- a/docs/api/segmentation/totalsegmentator.rst
+++ b/docs/api/segmentation/totalsegmentator.rst
@@ -2,6 +2,7 @@
 TotalSegmentator
 ================
 
+.. module:: physiomotion4d.segment_chest_total_segmentator
 .. currentmodule:: physiomotion4d
 
 ``SegmentChestTotalSegmentator`` groups a TotalSegmentator labelmap into the
@@ -39,10 +40,11 @@ Basic Usage
    itk.imwrite(vessels, "major_vessels_mask.nrrd")
    itk.imwrite(labelmap, "labelmap.nrrd")
 
-Returned Masks
-==============
+Returned Keys
+=============
 
-``segment()`` returns a dictionary with these keys:
+For this segmenter, ``segment()`` returns a dictionary with the following
+keys:
 
 * ``labelmap``
 * ``lung``
@@ -54,6 +56,11 @@ Returned Masks
 * ``contrast``
 
 The dictionary should be accessed by key. Do not unpack it positionally.
+The exact key set is determined by the segmenter's :class:`AnatomyTaxonomy`
+and may differ from other segmenters (see :doc:`base`). For
+:class:`SegmentChestTotalSegmentator` specifically, all seven groups plus
+``labelmap`` are always present; downstream code that targets a different
+segmenter should check membership.
 
 Operational Notes
 =================
diff --git a/docs/api/usd/index.rst b/docs/api/usd/index.rst
index 3be096a..d09a968 100644
--- a/docs/api/usd/index.rst
+++ b/docs/api/usd/index.rst
@@ -21,7 +21,8 @@ Quick Links
 **USD Modules**:
    * :doc:`tools` - Core USD utilities
    * :doc:`anatomy_tools` - Anatomical structure tools
-   * :doc:`vtk_conversion` - VTK to USD conversion
+   * :doc:`vtk_conversion` - VTK to USD conversion (preferred high-level API)
+   * :doc:`vtk_to_usd_lib` - Low-level ``vtk_to_usd`` subpackage (advanced)
 
 Module Documentation
 ====================
@@ -32,6 +33,7 @@ Module Documentation
    tools
    anatomy_tools
    vtk_conversion
+   vtk_to_usd_lib
 
 Quick Start
 ===========
diff --git a/docs/api/usd/vtk_conversion.rst b/docs/api/usd/vtk_conversion.rst
index d2d59f4..32f8f58 100644
--- a/docs/api/usd/vtk_conversion.rst
+++ b/docs/api/usd/vtk_conversion.rst
@@ -2,6 +2,7 @@
 VTK to USD Conversion
 ====================================
 
+.. module:: physiomotion4d.convert_vtk_to_usd
 .. currentmodule:: physiomotion4d
 
 Convert VTK mesh files to USD format with animation support.
diff --git a/docs/api/usd/vtk_to_usd_lib.rst b/docs/api/usd/vtk_to_usd_lib.rst
new file mode 100644
index 0000000..4bddc4e
--- /dev/null
+++ b/docs/api/usd/vtk_to_usd_lib.rst
@@ -0,0 +1,86 @@
+===============================
+Low-Level vtk_to_usd Subpackage
+===============================
+
+.. module:: physiomotion4d.vtk_to_usd
+.. currentmodule:: physiomotion4d.vtk_to_usd
+
+``physiomotion4d.vtk_to_usd`` is a stable low-level API for advanced external
+users. Inside this repository (experiments, workflows, CLIs, tutorials,
+tests), use :class:`~physiomotion4d.ConvertVTKToUSD` from
+:doc:`vtk_conversion` instead of importing this subpackage directly.
+
+This subpackage exposes the readers, data containers, coordinate helpers, and
+USD primitive writers that back ``ConvertVTKToUSD``. The public symbols are
+documented in their defining submodules below; ``__init__`` re-exports them
+for convenience.
+
+File Facade
+===========
+
+.. automodule:: physiomotion4d.vtk_to_usd.converter
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Data Structures
+===============
+
+.. automodule:: physiomotion4d.vtk_to_usd.data_structures
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+VTK Readers
+===========
+
+.. automodule:: physiomotion4d.vtk_to_usd.vtk_reader
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+USD Mesh Conversion
+===================
+
+.. automodule:: physiomotion4d.vtk_to_usd.usd_mesh_converter
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Material Manager
+================
+
+.. automodule:: physiomotion4d.vtk_to_usd.material_manager
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Mesh Utilities
+==============
+
+.. automodule:: physiomotion4d.vtk_to_usd.mesh_utils
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+USD Coordinate and Primvar Helpers
+==================================
+
+.. automodule:: physiomotion4d.vtk_to_usd.usd_utils
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+Primvar Derivations
+===================
+
+.. automodule:: physiomotion4d.vtk_to_usd.primvar_derivations
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+See Also
+========
+
+* :doc:`vtk_conversion`
+* :doc:`../workflows`
diff --git a/docs/api/utilities/contour_tools.rst b/docs/api/utilities/contour_tools.rst
index 3627116..3e4b881 100644
--- a/docs/api/utilities/contour_tools.rst
+++ b/docs/api/utilities/contour_tools.rst
@@ -15,4 +15,4 @@ Module Reference
 
 .. rubric:: Navigation
 
-:doc:`transform_tools` | :doc:`index` | :doc:`nrrd_conversion`
+:doc:`transform_tools` | :doc:`index` | :doc:`nrrd_conversion` | :doc:`test_tools`
diff --git a/docs/api/utilities/data_download.rst b/docs/api/utilities/data_download.rst
new file mode 100644
index 0000000..a84be55
--- /dev/null
+++ b/docs/api/utilities/data_download.rst
@@ -0,0 +1,20 @@
+==============
+Data Downloads
+==============
+
+.. currentmodule:: physiomotion4d
+
+Helpers for downloading and verifying optional example datasets used by the
+tutorials and experiments.
+
+Module Reference
+================
+
+.. automodule:: physiomotion4d.data_download_tools
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+.. rubric:: Navigation
+
+:doc:`test_tools` | :doc:`index`
diff --git a/docs/api/utilities/index.rst b/docs/api/utilities/index.rst
index 481084b..508ad35 100644
--- a/docs/api/utilities/index.rst
+++ b/docs/api/utilities/index.rst
@@ -24,6 +24,8 @@ Quick Links
    * :doc:`transform_tools` - Transform operations
    * :doc:`contour_tools` - Contour processing
    * :doc:`nrrd_conversion` - NRRD file utilities
+   * :doc:`test_tools` - Baseline / result comparison helpers
+   * :doc:`data_download` - Optional dataset download helpers
 
 Module Documentation
 ====================
@@ -35,6 +37,8 @@ Module Documentation
    transform_tools
    contour_tools
    nrrd_conversion
+   test_tools
+   data_download
 
 See Also
 ========
diff --git a/docs/api/utilities/test_tools.rst b/docs/api/utilities/test_tools.rst
new file mode 100644
index 0000000..534da67
--- /dev/null
+++ b/docs/api/utilities/test_tools.rst
@@ -0,0 +1,20 @@
+==========
+Test Tools
+==========
+
+.. currentmodule:: physiomotion4d
+
+Baseline vs. result image comparison utilities used by the PhysioMotion4D
+pytest suite.
+
+Module Reference
+================
+
+.. automodule:: physiomotion4d.test_tools
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+.. rubric:: Navigation
+
+:doc:`nrrd_conversion` | :doc:`index` | :doc:`data_download`
diff --git a/docs/api/workflows.rst b/docs/api/workflows.rst
index 1d70119..d3e5be2 100644
--- a/docs/api/workflows.rst
+++ b/docs/api/workflows.rst
@@ -2,6 +2,12 @@
 Workflow Classes
 ================
 
+.. module:: physiomotion4d.workflow_convert_heart_gated_ct_to_usd
+.. module:: physiomotion4d.workflow_convert_ct_to_vtk
+.. module:: physiomotion4d.workflow_convert_vtk_to_usd
+.. module:: physiomotion4d.workflow_create_statistical_model
+.. module:: physiomotion4d.workflow_fit_statistical_model_to_patient
+.. module:: physiomotion4d.workflow_reconstruct_highres_4d_ct
 .. currentmodule:: physiomotion4d
 
 Workflow classes are the highest-level Python API in PhysioMotion4D. They
diff --git a/docs/cli_scripts/create_statistical_model.rst b/docs/cli_scripts/create_statistical_model.rst
index ccd6af6..12e5e5b 100644
--- a/docs/cli_scripts/create_statistical_model.rst
+++ b/docs/cli_scripts/create_statistical_model.rst
@@ -8,7 +8,7 @@ Overview
 The ``physiomotion4d-create-statistical-model`` command-line tool builds a PCA
 (Principal Component Analysis) statistical shape model from a sample of meshes
 aligned to a reference mesh. This mirrors the pipeline in the
-Heart-Create_Statistical_Model experiment notebooks.
+Heart-Create_Statistical_Model experiment scripts.
 
 The workflow:
 
diff --git a/docs/contributing.rst b/docs/contributing.rst
index fb57858..f852503 100644
--- a/docs/contributing.rst
+++ b/docs/contributing.rst
@@ -37,6 +37,14 @@ Getting Started
 
       pip install -e ".[dev]"
 
+   To install the full developer environment with CUDA 13, documentation, test,
+   and development dependencies, use either of these equivalent uv forms:
+
+   .. code-block:: bash
+
+      uv pip install -e ".[cuda13,docs,test,dev]"
+      uv pip install -e . --all-extras
+
 5. **Install pre-commit hooks**:
 
    .. code-block:: bash
@@ -64,10 +72,10 @@ For the best development experience with VS Code or Cursor, install these extens
 
 **Not Needed (Replaced by Ruff):**
 
-* ❌ ms-python.black-formatter - No longer needed
-* ❌ ms-python.isort - No longer needed
-* ❌ ms-python.flake8 - No longer needed
-* ❌ ms-python.pylint - No longer needed
+* ms-python.black-formatter - No longer needed
+* ms-python.isort - No longer needed
+* ms-python.flake8 - No longer needed
+* ms-python.pylint - No longer needed
 
 VS Code Settings
 ----------------
@@ -101,9 +109,10 @@ This configuration:
 Experiment Scripts
 ------------------
 
-Experiments in the ``experiments/`` directory are ``# %%`` percent-format Python
-scripts (Jupytext style). They run interactively via the VS Code Python extension's
-"Run Cell" feature — no Jupyter extension required.
+Experiments in the ``experiments/`` directory are ``# %%`` percent-format
+Python scripts. They run end-to-end as plain Python (``python <script>.py``)
+or cell-by-cell via the VS Code / Cursor Python extension's "Run Cell"
+feature.
 
 * Ruff formats these cell-separated scripts automatically
 * Type checking is less strict in experiment scripts (expected for exploratory work)
@@ -113,13 +122,16 @@ First-Time Setup Checklist
 
 After cloning the repository:
 
-1. ✅ Install Python 3.10+ and create virtual environment
-2. ✅ Install development dependencies: ``pip install -e ".[dev]"``
-3. ✅ Install pre-commit hooks: ``pre-commit install``
-4. ✅ Install Ruff extension in VS Code/Cursor
-5. ✅ Remove old formatter extensions (black, isort, flake8, pylint)
-6. ✅ Verify settings: Open a Python file and save to test auto-formatting
-7. ✅ Run tests: ``pytest tests/ -m "not slow"`` to verify setup
+1. Install Python 3.11+ and create virtual environment
+2. Install development dependencies: ``pip install -e ".[dev]"``,
+   or install all extras with
+   ``uv pip install -e ".[cuda13,docs,test,dev]"`` or
+   ``uv pip install -e . --all-extras``
+3. Install pre-commit hooks: ``pre-commit install``
+4. Install Ruff extension in VS Code/Cursor
+5. Remove old formatter extensions (black, isort, flake8, pylint)
+6. Verify settings: Open a Python file and save to test auto-formatting
+7. Run tests: ``pytest tests/ -m "not slow"`` to verify setup
 
 Code Style
 ==========
@@ -289,7 +301,7 @@ When contributing new workflows or examples:
 
 **Production Code (src/physiomotion4d/cli/):**
 
-* ✅ **DO contribute here** for production-ready CLI implementations
+* **DO contribute here** for production-ready CLI implementations
 * Must include proper error handling and validation
 * Should follow all code style and testing requirements
 * Serves as definitive usage examples for users
@@ -297,16 +309,12 @@ When contributing new workflows or examples:
 
 **Research Code (experiments/ directory):**
 
-* 💡 **May contribute here** for exploratory research and design experiments
+* **May contribute here** for exploratory research and design experiments
 * Can have hardcoded paths and minimal error handling
 * Should document what was learned and how it informed production code
 * Helps others understand adaptation possibilities for new domains
 * Should reference corresponding production implementation in CLI commands or ``src/physiomotion4d/cli/``
 
-**Key Principle:** If users might copy your code for production use, it should use the CLI commands
-or extend the implementations in ``src/physiomotion4d/cli/``.
-If it's a proof-of-concept demonstrating what's possible, it belongs in ``experiments/``.
-
 Docstring Format
 ----------------
 
diff --git a/docs/developer/segmentation.rst b/docs/developer/segmentation.rst
index 3fe56bd..a36b50c 100644
--- a/docs/developer/segmentation.rst
+++ b/docs/developer/segmentation.rst
@@ -18,11 +18,14 @@ Current Segmentation Contract
    segmenter = SegmentChestTotalSegmentator()
    masks = segmenter.segment(image, contrast_enhanced_study=True)
 
-   heart = masks["heart"]
    labelmap = masks["labelmap"]
+   if "heart" in masks:
+       heart = masks["heart"]
 
-Segmentation outputs are dictionaries of ITK images. Access masks by key rather
-than by positional unpacking.
+Segmentation outputs are dictionaries of ITK images. Access masks by key,
+not by positional unpacking. The exact key set depends on the segmenter's
+:class:`physiomotion4d.AnatomyTaxonomy` — see :doc:`../api/segmentation/base`
+for the per-segmenter key sets and the general contract.
 
 Implemented Segmenters
 ======================
@@ -31,12 +34,73 @@ Implemented Segmenters
 * :class:`physiomotion4d.SegmentHeartSimpleware`
 * :class:`physiomotion4d.SegmentAnatomyBase`
 
+Adding a New Segmenter
+======================
+
+A new segmenter subclass declares which anatomy groups and organ labels it
+produces by populating ``self.taxonomy``. The base class owns the
+:class:`AnatomyTaxonomy` instance and contributes two default placeholders
+(``contrast`` at id 135, ``soft_tissue`` at id 133).
+
+.. code-block:: python
+
+   import logging
+
+   from physiomotion4d import SegmentAnatomyBase
+
+
+   class SegmentMySite(SegmentAnatomyBase):
+       def __init__(self, log_level=logging.INFO):
+           super().__init__(log_level=log_level)
+           self.target_spacing = 1.5
+
+           # Register organ labels under group names. New group names
+           # can be introduced freely; downstream renderers fall back
+           # to render_params["other"] for unregistered groups.
+           for group_name, organs in (
+               ("heart", {51: "heart", 61: "atrial_appendage_left"}),
+               ("lung", {10: "lung_upper_lobe_left"}),
+           ):
+               for label_id, organ_name in organs.items():
+                   self.taxonomy.add_organ(group_name, label_id, organ_name)
+
+           # Fill in 'other' for any unclaimed ids in the [1, 256) range.
+           self._finalize_other_group()
+
+       def segmentation_method(self, preprocessed_image):
+           # Run the actual model here; must return an itk.Image labelmap.
+           ...
+
+The base class then provides the shared pre/post-processing, contrast-agent
+fusion, and ``create_anatomy_group_masks()`` that walks the taxonomy and
+emits one mask per registered group plus ``"other"``.
+
+Custom Anatomy Looks
+====================
+
+If the new segmenter introduces groups beyond the default chest set
+(``heart``, ``lung``, ``bone``, ``major_vessels``, ``contrast``,
+``soft_tissue``, ``other``), register a matching OmniSurface look so the
+USD renderer doesn't fall back to the generic ``"other"`` material:
+
+.. code-block:: python
+
+   from physiomotion4d.usd_anatomy_tools import DEFAULT_RENDER_PARAMS
+
+   DEFAULT_RENDER_PARAMS["brain"] = {
+       "name": "Brain",
+       "diffuse_reflection_color": (0.85, 0.75, 0.7),
+       # ... see existing entries for the full parameter list ...
+   }
+
+See :doc:`usd_generation` for the renderer-side contract.
+
 Development Notes
 =================
 
-* New runtime segmenters should inherit from ``SegmentAnatomyBase`` or
-  ``PhysioMotion4DBase``.
-* Use ``log_info()`` and ``log_debug()`` inside runtime classes.
+* Use ``log_info()`` and ``log_debug()`` inside runtime classes; never ``print``.
+* Document the key set the segmenter produces; downstream callers should
+  check membership rather than assume a fixed schema.
 * Keep tests synthetic unless real model/data behavior is being validated.
 * Mark real-data tests with ``requires_data``.
 
@@ -44,4 +108,6 @@ See Also
 ========
 
 * :doc:`../api/segmentation/index`
+* :doc:`../api/segmentation/base` — full AnatomyTaxonomy reference
 * :doc:`workflows`
+* :doc:`usd_generation` — how the taxonomy drives USD output
diff --git a/docs/developer/usd_generation.rst b/docs/developer/usd_generation.rst
index 742bbf0..4d0375a 100644
--- a/docs/developer/usd_generation.rst
+++ b/docs/developer/usd_generation.rst
@@ -60,6 +60,48 @@ In-Memory Meshes
    )
    stage = converter.convert('cardiac_model.usd')
 
+Labeled Meshes with Anatomy Grouping
+------------------------------------
+
+When the input has anatomical labels (``mask_ids`` and a ``boundary_labels``
+cell array), pass a :class:`physiomotion4d.SegmentAnatomyBase` instance
+through the ``segmenter`` argument so labeled prims are grouped by anatomy
+type. The output layout becomes:
+
+* Meshes: ``/World/{basename}/{group}/{organ_name}``
+* Materials: ``/World/Looks/{group}/{organ_name}_material``
+
+with one Xform per group (``heart``, ``lung``, ``bone``, ...) that only
+appears if at least one labeled organ from the input falls into that group.
+
+.. code-block:: python
+
+   from physiomotion4d import (
+       ConvertVTKToUSD,
+       SegmentChestTotalSegmentator,
+   )
+
+   seg = SegmentChestTotalSegmentator()
+   converter = ConvertVTKToUSD(
+       data_basename='Patient',
+       input_polydata=meshes,
+       mask_ids=seg.taxonomy.all_labels(),
+       segmenter=seg,
+   )
+   stage = converter.convert('patient.usd')
+   # /World/Patient/heart/heart, /World/Patient/lung/lung_upper_lobe_left, ...
+   # /World/Looks/heart/heart_material, ...
+
+If no segmenter is supplied, labels are grouped under a single ``Anatomy``
+Xform (``/World/{basename}/Anatomy/{organ_name}``) so the hierarchy stays
+two-deep regardless.
+
+The classification logic is delegated to the segmenter's
+:class:`AnatomyTaxonomy`; new groups added there flow into the prim layout
+without any change to ``ConvertVTKToUSD``. See
+:doc:`../api/segmentation/base` for the taxonomy contract and
+:doc:`segmentation` for how to add a new segmenter.
+
 Colormaps
 ---------
 
@@ -72,6 +114,81 @@ Colormaps
    )
    stage = converter.convert('pressure.usd')
 
+Derived Scalars: von Mises Stress
+---------------------------------
+
+For finite-element outputs that carry a 9-component stress tensor field
+(e.g. cardiac valve simulations), call ``compute_von_mises_stress`` between
+``from_files`` and ``convert`` to add a scalar ``von_mises_stress`` primvar
+suitable for colormap rendering:
+
+.. code-block:: python
+
+   stage = (
+       ConvertVTKToUSD.from_files(
+           data_basename='Valve',
+           vtk_files=valve_files,
+           time_codes=time_codes,
+           separate_by='connectivity',
+       )
+       .compute_von_mises_stress('stress')   # source array name in the VTK
+       .convert('valve.usd')
+   )
+
+The default output array name is ``"von_mises_stress"``; the method finds
+the source array in either ``point_data`` or ``cell_data`` and writes the
+scalar back to the same data dict so the convert step picks it up as a
+USD primvar.
+
+Framing Camera
+--------------
+
+Every USD stage that ``ConvertVTKToUSD`` (and the lower-level
+``convert_vtk_file`` facade) writes gets a ``/World/Camera`` prim with a
+tight ``clippingRange`` sized to the geometry's bounding box. This avoids
+the common "Omniverse Kit near-plane clips small geometry" problem for
+medical-scale meshes (~0.03 m wide). The camera is also baked into stages
+produced by ``TransformTools.convert_transform_to_usd_visualization`` and
+``USDTools.merge_usd_files``; the helper is idempotent so re-merging a USD
+that already has a Camera does not produce a duplicate transform op.
+
+Anatomy Materials with USDAnatomyTools
+=======================================
+
+:class:`physiomotion4d.USDAnatomyTools` applies OmniSurface materials to
+labeled meshes after conversion. It reads :class:`AnatomyTaxonomy` from the
+segmenter to find which prim names map to which group, and looks up the
+material parameters in its ``render_params`` dict (initialized from the
+module-level :data:`physiomotion4d.usd_anatomy_tools.DEFAULT_RENDER_PARAMS`).
+
+.. code-block:: python
+
+   from physiomotion4d import USDAnatomyTools
+
+   tools = USDAnatomyTools(stage)
+   tools.enhance_meshes(seg)
+   stage.Export('painted.usd')
+
+Add a custom anatomy look without subclassing by mutating either the
+module-level defaults (affects future instances) or a single instance:
+
+.. code-block:: python
+
+   from physiomotion4d.usd_anatomy_tools import DEFAULT_RENDER_PARAMS
+
+   DEFAULT_RENDER_PARAMS["brain"] = {
+       "name": "Brain",
+       "diffuse_reflection_color": (0.85, 0.75, 0.7),
+       # ... full parameter list mirrors the existing entries ...
+   }
+
+   # ...or per-instance:
+   tools.render_params["brain"] = {...}
+
+``enhance_meshes`` falls back to ``render_params["other"]`` for any group
+without a registered entry, so newly registered groups still render — just
+with the generic "other" look until a dedicated entry is added.
+
 Advanced Low-Level Facade
 =========================
 
diff --git a/docs/examples.rst b/docs/examples.rst
index 8653759..d1cd0cb 100644
--- a/docs/examples.rst
+++ b/docs/examples.rst
@@ -416,9 +416,9 @@ Batch process multiple datasets:
 
        try:
            final_usd = workflow.process()
-           print(f"  ✓ Complete: {final_usd}")
+           print(f"  Complete: {final_usd}")
        except Exception as e:
-           print(f"  ✗ Failed: {e}")
+           print(f"  Failed: {e}")
 
 Parallel Segmentation
 ---------------------
@@ -459,14 +459,10 @@ Download Slicer-Heart Dataset
 
 .. code-block:: python
 
-   from pathlib import Path
-   from urllib.request import urlretrieve
-
-   data_dir = Path("data/Slicer-Heart-CT")
-   data_dir.mkdir(parents=True, exist_ok=True)
+   from physiomotion4d import DataDownloadTools
 
-   url = "https://github.com/SlicerHeart/SlicerHeart/releases/download/TestingData/TruncalValve_4DCT.seq.nrrd"
-   urlretrieve(url, data_dir / "TruncalValve_4DCT.seq.nrrd")
+   data_file = DataDownloadTools.DownloadSlicerHeartCTData("data/Slicer-Heart-CT")
+   assert DataDownloadTools.VerifySlicerHeartCTData("data/Slicer-Heart-CT")
 
 Custom Workflow Examples
 =========================
diff --git a/docs/faq.rst b/docs/faq.rst
index 1bd0b6b..617c542 100644
--- a/docs/faq.rst
+++ b/docs/faq.rst
@@ -63,7 +63,7 @@ torchvision, and torchaudio are sourced from
 What Python version is required?
 ---------------------------------
 
-Python 3.10, 3.11, or 3.12 are supported.
+Python 3.11 or 3.12 are supported.
 
 Usage Questions
 ===============
diff --git a/docs/quickstart.rst b/docs/quickstart.rst
index 074d3a6..ef29a4a 100644
--- a/docs/quickstart.rst
+++ b/docs/quickstart.rst
@@ -15,25 +15,24 @@ This guide will help you get started with PhysioMotion4D quickly.
 Tutorials
 =========
 
-The ``tutorials/`` directory contains six end-to-end scripts, one per major
-workflow.  Each script is self-contained, includes its own ``argparse`` CLI, and
-can be imported as a module from the test suite.
+The ``tutorials/`` directory contains nine end-to-end scripts, one per major
+workflow. Each script is a ``# %%`` percent-cell Python script that exercises
+the workflow classes directly. Run as a regular file
+(``python tutorials/tutorial_01_...py``) or cell-by-cell in VS Code or Cursor.
 
 See :doc:`tutorials` for the NVIDIA-styled tutorial card index, dataset
-requirements, commands, and workflow details.
+requirements, script paths, and workflow details.
 
 After preparing the Slicer-Heart-CT data, run the first two tutorials:
 
 .. code-block:: bash
 
-   python tutorials/tutorial_01_heart_gated_ct_to_usd.py \
-       --data-dir ./data --output-dir ./output/tutorial_01 \
-       --registration-method ants
+   python tutorials/tutorial_01_heart_gated_ct_to_usd.py
 
-   python tutorials/tutorial_02_ct_to_vtk.py \
-       --data-dir ./data --output-dir ./output/tutorial_02
+   python tutorials/tutorial_02_ct_to_vtk.py
 
-Each script prints the paths of outputs and screenshots it created.
+Tutorial paths are defined near the top of each script. To use different paths,
+edit the script constants or use the installed ``physiomotion4d-*`` CLI commands.
 See ``tutorials/README.md`` for dataset download instructions and the
 recommended run order.
 
@@ -44,6 +43,9 @@ Recommended run order:
 3. Tutorial 3 after downloading KCL-Heart-Model.
 4. Tutorial 4 after Tutorial 3 because it can consume Tutorial 3 output.
 5. Tutorial 6 after downloading DirLab-4DCT.
+6. Tutorial 7 after downloading DirLab-4DCT.
+7. Tutorial 8 after Tutorial 7 because it consumes the fitted PCA meshes.
+8. Tutorial 9 after Tutorial 8 because it trains from propagated meshes.
 
 Prerequisites
 =============
@@ -66,7 +68,7 @@ CUDA-capable GPU are required for practical runtime.
 
 .. code-block:: bash
 
-   python -c "import pathlib, urllib.request; pathlib.Path('data/test').mkdir(parents=True, exist_ok=True); urllib.request.urlretrieve('https://github.com/SlicerHeart/SlicerHeart/releases/download/TestingData/TruncalValve_4DCT.seq.nrrd', 'data/test/TruncalValve_4DCT.seq.nrrd')"
+   python -c "from physiomotion4d import DataDownloadTools; DataDownloadTools.DownloadSlicerHeartCTData('data/test')"
 
    physiomotion4d-heart-gated-ct data/test/TruncalValve_4DCT.seq.nrrd \
        --registration-method ants \
@@ -237,18 +239,15 @@ Download Sample Cardiac CT Data
 
 .. code-block:: python
 
-   import urllib.request
-   import os
+   from physiomotion4d import DataDownloadTools
 
-   # Create data directory
-   os.makedirs("sample_data", exist_ok=True)
-
-   # Download sample from Slicer-Heart-CT
-   url = "https://github.com/SlicerHeart/SlicerHeart/releases/download/TestingData/TruncalValve_4DCT.seq.nrrd"
-   urllib.request.urlretrieve(url, "sample_data/TruncalValve_4DCT.seq.nrrd")
+   data_file = DataDownloadTools.DownloadSlicerHeartCTData("sample_data")
+   assert DataDownloadTools.VerifySlicerHeartCTData("sample_data")
 
 DirLab-4DCT data is manual-only; see ``data/README.md`` before running the
-high-resolution 4D CT reconstruction tutorial.
+high-resolution 4D CT reconstruction, lung-lobe PCA model, or PCA time-series
+propagation tutorials. Tutorial 9 also requires the PhysicsNeMo dependency
+installed with PhysioMotion4D.
 
 Visualizing Results
 ===================
diff --git a/docs/tutorials.rst b/docs/tutorials.rst
index 9808dd7..2e5313b 100644
--- a/docs/tutorials.rst
+++ b/docs/tutorials.rst
@@ -11,10 +11,10 @@ Tutorials
      <p class="pm4d-kicker">PhysioMotion4D tutorials</p>
      <h1>Build animated medical USD workflows for NVIDIA Omniverse</h1>
      <p>
-       Six focused tutorials walk through CT segmentation, registration,
+       Nine focused tutorials walk through CT segmentation, registration,
        statistical model fitting, high-resolution 4D reconstruction, and USD
        export. Each card links to implementation details, datasets, and the
-       command used to run the workflow.
+       percent-cell Python script used to run the workflow.
      </p>
    </section>
 
@@ -55,16 +55,43 @@ Tutorials
        <p>Register respiratory CT phases and reconstruct a higher-resolution 4D volume series.</p>
        <span class="pm4d-card__meta">DirLab-4DCT</span>
      </a>
+     <a class="pm4d-card" href="#tutorial-7-dirlab-lung-lobe-pca-model">
+       <span class="pm4d-card__number">07</span>
+       <h2>DirLab Lung-Lobe PCA Model</h2>
+       <p>Build a surface PCA model from five lung lobes and fit it to all cases.</p>
+       <span class="pm4d-card__meta">DirLab-4DCT</span>
+     </a>
+     <a class="pm4d-card" href="#tutorial-8-dirlab-pca-time-series-propagation">
+       <span class="pm4d-card__number">08</span>
+       <h2>DirLab PCA Time-Series Propagation</h2>
+       <p>Register respiratory phases with ANTs+ICON and propagate fitted meshes.</p>
+       <span class="pm4d-card__meta">Tutorial 7 output</span>
+     </a>
+     <a class="pm4d-card" href="#tutorial-9-physicsnemo-mesh-stage-model">
+       <span class="pm4d-card__number">09</span>
+       <h2>PhysicsNeMo Mesh Stage Model</h2>
+       <p>Train a PhysicsNeMo MLP to predict lung-lobe meshes at requested stages.</p>
+       <span class="pm4d-card__meta">Tutorial 8 output</span>
+     </a>
    </section>
 
 Recommended Run Order
 =====================
 
+Tutorials are ``# %%`` percent-cell Python scripts. Each script defines its
+data and output paths near the top, using repository ``data/`` and ``output/``
+directories by default. Edit those constants for tutorial exploration, or use
+the installed ``physiomotion4d-*`` CLI commands when you need command-line path
+arguments.
+
 1. Run Tutorials 1 and 2 after preparing Slicer-Heart-CT data.
 2. Run Tutorial 5 after Tutorial 2 because it consumes Tutorial 2 output.
 3. Run Tutorial 3 after downloading KCL-Heart-Model.
 4. Run Tutorial 4 after Tutorial 3 because it can consume the PCA model output.
 5. Run Tutorial 6 after downloading DirLab-4DCT.
+6. Run Tutorial 7 after downloading DirLab-4DCT.
+7. Run Tutorial 8 after Tutorial 7 because it consumes fitted PCA meshes.
+8. Run Tutorial 9 after Tutorial 8 because it trains from propagated meshes.
 
 Tutorial 1: Heart-Gated CT to Animated USD
 ==========================================
@@ -78,12 +105,10 @@ Workflow
 Dataset
    Slicer-Heart-CT, prepared before running the tutorial.
 
-Command
+Run
    .. code-block:: bash
 
-      python tutorials/tutorial_01_heart_gated_ct_to_usd.py \
-          --data-dir ./data --output-dir ./output/tutorial_01 \
-          --registration-method ants
+      python tutorials/tutorial_01_heart_gated_ct_to_usd.py
 
 Outputs
    Registered phase images, transformed contours, preview screenshots, and an
@@ -101,11 +126,10 @@ Workflow
 Dataset
    Slicer-Heart-CT, prepared before running the tutorial.
 
-Command
+Run
    .. code-block:: bash
 
-      python tutorials/tutorial_02_ct_to_vtk.py \
-          --data-dir ./data --output-dir ./output/tutorial_02
+      python tutorials/tutorial_02_ct_to_vtk.py
 
 Outputs
    Segmentation artifacts, VTK PolyData surfaces, and preview screenshots.
@@ -122,11 +146,10 @@ Workflow
 Dataset
    KCL-Heart-Model, downloaded manually.
 
-Command
+Run
    .. code-block:: bash
 
-      python tutorials/tutorial_03_create_statistical_model.py \
-          --data-dir ./data --output-dir ./output/tutorial_03
+      python tutorials/tutorial_03_create_statistical_model.py
 
 Outputs
    PCA model files, mean shape, and component diagnostics.
@@ -143,12 +166,10 @@ Workflow
 Dataset
    KCL-Heart-Model, downloaded manually.
 
-Command
+Run
    .. code-block:: bash
 
-      python tutorials/tutorial_04_fit_statistical_model_to_patient.py \
-          --data-dir ./data --output-dir ./output/tutorial_04 \
-          --pca-json ./output/tutorial_03/pca_model.json
+      python tutorials/tutorial_04_fit_statistical_model_to_patient.py
 
 Outputs
    Patient-fitted statistical model surfaces and registration diagnostics.
@@ -165,12 +186,10 @@ Workflow
 Dataset
    Output from Tutorial 2.
 
-Command
+Run
    .. code-block:: bash
 
-      python tutorials/tutorial_05_vtk_to_usd.py \
-          --data-dir ./data --output-dir ./output/tutorial_05 \
-          --vtk-file output/tutorial_02/patient_surfaces.vtp
+      python tutorials/tutorial_05_vtk_to_usd.py
 
 Outputs
    Time-sampled USD scene and conversion logs for Omniverse inspection.
@@ -187,16 +206,80 @@ Workflow
 Dataset
    DirLab-4DCT, downloaded manually.
 
-Command
+Run
    .. code-block:: bash
 
-      python tutorials/tutorial_06_reconstruct_highres_4d_ct.py \
-          --data-dir ./data --output-dir ./output/tutorial_06
+      python tutorials/tutorial_06_reconstruct_highres_4d_ct.py
 
 Outputs
    Registered respiratory phases, reconstructed high-resolution CT volumes,
    and preview screenshots.
 
+Tutorial 7: DirLab Lung-Lobe PCA Model
+======================================
+
+Script
+   ``tutorials/tutorial_07_dirlab_pca_model.py``
+
+Workflow
+   ``WorkflowConvertCTToVTK``, ``WorkflowCreateStatisticalModel``, and
+   ``WorkflowFitStatisticalModelToPatient``
+
+Dataset
+   DirLab-4DCT, downloaded manually.
+
+Run
+   .. code-block:: bash
+
+      python tutorials/tutorial_07_dirlab_pca_model.py
+
+Outputs
+   Five-lobe lung surface meshes, a surface PCA model, and PCA-fitted surfaces
+   for every available case.
+
+Tutorial 8: DirLab PCA Time-Series Propagation
+==============================================
+
+Script
+   ``tutorials/tutorial_08_dirlab_pca_time_series.py``
+
+Workflow
+   ``RegisterTimeSeriesImages`` with ``registration_method='ants_icon'`` and
+   ``TransformTools``
+
+Dataset
+   DirLab-4DCT plus Tutorial 7 fitted mesh outputs.
+
+Run
+   .. code-block:: bash
+
+      python tutorials/tutorial_08_dirlab_pca_time_series.py
+
+Outputs
+   Per-case ANTs+ICON transforms and one PCA-fitted lung-lobe surface for each
+   DirLab respiratory phase.
+
+Tutorial 9: PhysicsNeMo Mesh Stage Model
+========================================
+
+Script
+   ``tutorials/tutorial_09_physicsnemo_mesh_stage_model.py``
+
+Workflow
+   ``physicsnemo.models.mlp.FullyConnected`` trained on Tutorial 8 meshes.
+
+Dataset
+   Tutorial 8 propagated PCA mesh outputs.
+
+Run
+   .. code-block:: bash
+
+      python tutorials/tutorial_09_physicsnemo_mesh_stage_model.py
+
+Outputs
+   Per-case PhysicsNeMo checkpoints, training metadata, loss histories, and a
+   predicted PCA-fitted mesh at the requested normalized respiratory stage.
+
 Dataset Notes
 =============
 
diff --git a/experiments/Colormap-VTK_To_USD/colormap_vtk_to_usd.py b/experiments/Colormap-VTK_To_USD/colormap_vtk_to_usd.py
index 8d2af62..15b8f29 100644
--- a/experiments/Colormap-VTK_To_USD/colormap_vtk_to_usd.py
+++ b/experiments/Colormap-VTK_To_USD/colormap_vtk_to_usd.py
@@ -131,6 +131,7 @@ def create_example_mesh_with_data(time_step):
     data_basename="CardiacModel", input_polydata=meshes, mask_ids=None
 )
 
+print("Setting colormap")
 converter.set_colormap(
     color_by_array="transmembrane_potential",
     colormap="rainbow",
@@ -138,6 +139,7 @@ def create_example_mesh_with_data(time_step):
 )
 
 output_file = output_dir / "example2_rainbow_custom.usd"
+print("Creating file:", output_file)
 stage = converter.convert(str(output_file))
 print(f"✓ Created: {output_file}")
 
diff --git a/experiments/Convert_VTK_To_USD/convert_chop_alterra_valve_to_usd.py b/experiments/Convert_VTK_To_USD/convert_chop_alterra_valve_to_usd.py
index 8bbeb19..c05915c 100644
--- a/experiments/Convert_VTK_To_USD/convert_chop_alterra_valve_to_usd.py
+++ b/experiments/Convert_VTK_To_USD/convert_chop_alterra_valve_to_usd.py
@@ -29,34 +29,36 @@
 import re
 from pathlib import Path
 
+from physiomotion4d import ConvertVTKToUSD
+
 # Use as a test
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 
 # Import USDTools for post-processing colormap
 from physiomotion4d.usd_tools import USDTools
 
-from physiomotion4d import ConvertVTKToUSD
-
 # %% [markdown]
 # ## 1. Discover and Organize Time-Series Files
 
 # %%
 # Set to True to use as a test.  Automatically done by
-#    running_as_test() helper function.
-quick_run = running_as_test()
-quick_run_step = 4
-
-# Define data directories (Alterra only)
-data_dir = Path.cwd().parent.parent / "data" / "CHOP-Valve4D"
+#    TestTools.running_as_test() helper function.
+test_mode = TestTools.running_as_test()
+test_mode_step = 4
+
+# Define data directories (Alterra only). Anchored to the script's location
+# so the experiment runs from any working directory.
+script_dir = Path(__file__).resolve().parent
+data_dir = script_dir.parent.parent / "data" / "CHOP-Valve4D"
 alterra_dir = data_dir / "Alterra"
 
-output_dir = Path.cwd() / "results" / "valve4d-alterra"
-if quick_run:
-    output_usd = output_dir / "alterra_quick.usd"
+output_dir = script_dir / "results" / "valve4d-alterra"
+if test_mode:
+    output_usd = output_dir / "alterra_test.usd"
 else:
     output_usd = output_dir / "alterra_full.usd"
 
-colormap_primvar_substrs = ["stress", "strain"]
+colormap_primvar_substrs = ["von_mises_stress"]
 colormap_name = "jet"  # matplotlib colormap name
 colormap_range_min = 25
 colormap_range_max = 200
@@ -79,6 +81,7 @@
     if match:
         time_step = int(match.group(1))
         alterra_series.append((time_step, vtk_file))
+        print(f"Found file: {time_step:04d}: {vtk_file.name}")
 
 # Sort by time step
 alterra_series.sort(key=lambda x: x[0])
@@ -129,50 +132,43 @@
     )
 
 # %% [markdown]
-# ## 3. Convert TPV25
+# ## 3. Convert Alterra
 
 # %%
 alterra_files = [file_path for _, file_path in alterra_series]
 alterra_times = [float(time_step) for time_step, _ in alterra_series]
 
-if quick_run:
-    alterra_files = alterra_files[::quick_run_step]
-    alterra_times = alterra_times[::quick_run_step]
+if test_mode:
+    alterra_files = alterra_files[::test_mode_step]
+    alterra_times = alterra_times[::test_mode_step]
 
 print(f"\nConverting to: {output_usd}")
 print(f"Number of time steps: {len(alterra_times)}")
 print("\nThis may take several minutes...\n")
 
 # topology validation and conversion happen inside from_files()
-stage = ConvertVTKToUSD.from_files(
-    data_basename="AlterraValve",
-    vtk_files=alterra_files,
-    extract_surface=True,
-    separate_by=separate_by,
-    times_per_second=times_per_second,
-    solid_color=solid_color,
-    time_codes=alterra_times,
-).convert(str(output_usd))
+stage = (
+    ConvertVTKToUSD.from_files(
+        data_basename="AlterraValve",
+        vtk_files=alterra_files,
+        extract_surface=True,
+        separate_by=separate_by,
+        times_per_second=times_per_second,
+        solid_color=solid_color,
+        time_codes=alterra_times,
+    )
+    .compute_von_mises_stress("stress")
+    .convert(str(output_usd))
+)
 
 # %%
 usd_tools = USDTools()
 # ConvertVTKToUSD places prims at /World/{basename}/{part_name}.
-# Discover the target prim dynamically so the path stays valid regardless
-# of how many connected components or cell types the VTK file produces.
+vessel_paths = usd_tools.list_mesh_paths_under(stage, parent_path="/World/AlterraValve")
 if separate_by == "connectivity":
-    mesh_paths = usd_tools.list_mesh_paths_under(
-        stage, parent_path="/World/AlterraValve"
-    )
-    candidates = [
-        p for p in mesh_paths if p.split("/")[-1].startswith("AlterraValve_object")
-    ]
-    vessel_path = candidates[-1] if candidates else "/World/AlterraValve/Mesh"
+    vessel_path = "/World/AlterraValve/AlterraValve_object3"
 elif separate_by == "cell_type":
-    mesh_paths = usd_tools.list_mesh_paths_under(
-        stage, parent_path="/World/AlterraValve"
-    )
-    triangle_paths = [p for p in mesh_paths if p.split("/")[-1].endswith("_Triangle")]
-    vessel_path = triangle_paths[0] if triangle_paths else "/World/AlterraValve/Mesh"
+    vessel_path = "/World/AlterraValve/AlterraValve_Triangle"
 else:
     vessel_path = "/World/AlterraValve/Mesh"
 
@@ -189,8 +185,8 @@
         str(output_usd),
         vessel_path,
         color_primvar,
-        intensity_range=(colormap_range_min, colormap_range_max),
         cmap=colormap_name,
+        intensity_range=(colormap_range_min, colormap_range_max),
         use_sigmoid_scale=True,
         bind_vertex_color_material=True,
     )
diff --git a/experiments/Convert_VTK_To_USD/convert_chop_tpv25_valve_to_usd.py b/experiments/Convert_VTK_To_USD/convert_chop_tpv25_valve_to_usd.py
index 64379a5..5722516 100644
--- a/experiments/Convert_VTK_To_USD/convert_chop_tpv25_valve_to_usd.py
+++ b/experiments/Convert_VTK_To_USD/convert_chop_tpv25_valve_to_usd.py
@@ -29,33 +29,34 @@
 import re
 from pathlib import Path
 
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d import ConvertVTKToUSD
+from physiomotion4d.test_tools import TestTools
 
 # Import USDTools for post-processing colormap
 from physiomotion4d.usd_tools import USDTools
 
-from physiomotion4d import ConvertVTKToUSD
-
 # %% [markdown]
 # ## 1. Discover and Organize Time-Series Files
 
 # %%
 # Set to True to use as a test.  Automatically done by
-#    running_as_test() helper function.
-quick_run = running_as_test()
+#    TestTools.running_as_test() helper function.
+quick_run = TestTools.running_as_test()
 quick_run_step = 4
 
-# Define data directories (TPV25 only)
-data_dir = Path.cwd().parent.parent / "data" / "CHOP-Valve4D"
+# Define data directories (TPV25 only). Anchored to the script's location
+# so the experiment runs from any working directory.
+script_dir = Path(__file__).resolve().parent
+data_dir = script_dir.parent.parent / "data" / "CHOP-Valve4D"
 tpv25_dir = data_dir / "TPV25"
 
-output_dir = Path.cwd() / "results" / "valve4d-tpv25"
+output_dir = script_dir / "results" / "valve4d-tpv25"
 if quick_run:
     output_usd = output_dir / "tpv25_quick.usd"
 else:
     output_usd = output_dir / "tpv25_full.usd"
 
-colormap_primvar_substrs = ["stress", "strain"]
+colormap_primvar_substrs = ["von_mises_stress"]
 colormap_name = "jet"  # matplotlib colormap name
 colormap_range_min = 25
 colormap_range_max = 200
@@ -143,31 +144,28 @@
 print("\nThis may take several minutes...\n")
 
 # topology validation and conversion happen inside from_files()
-stage = ConvertVTKToUSD.from_files(
-    data_basename="TPV25Valve",
-    vtk_files=tpv25_files,
-    extract_surface=True,
-    separate_by=separate_by,
-    times_per_second=times_per_second,
-    solid_color=solid_color,
-    time_codes=tpv25_times,
-).convert(str(output_usd))
+stage = (
+    ConvertVTKToUSD.from_files(
+        data_basename="TPV25Valve",
+        vtk_files=tpv25_files,
+        extract_surface=True,
+        separate_by=separate_by,
+        times_per_second=times_per_second,
+        solid_color=solid_color,
+        time_codes=tpv25_times,
+    )
+    .compute_von_mises_stress("stress")
+    .convert(str(output_usd))
+)
 
 # %%
 usd_tools = USDTools()
 # ConvertVTKToUSD places prims at /World/{basename}/{part_name}.
-# Discover the target prim dynamically so the path stays valid regardless
-# of how many connected components or cell types the VTK file produces.
+vessel_paths = usd_tools.list_mesh_paths_under(stage, parent_path="/World/TPV25Valve")
 if separate_by == "connectivity":
-    mesh_paths = usd_tools.list_mesh_paths_under(stage, parent_path="/World/TPV25Valve")
-    candidates = [
-        p for p in mesh_paths if p.split("/")[-1].startswith("TPV25Valve_object")
-    ]
-    vessel_path = candidates[-1] if candidates else "/World/TPV25Valve/Mesh"
+    vessel_path = "/World/TPV25Valve/TPV25Valve_object4"
 elif separate_by == "cell_type":
-    mesh_paths = usd_tools.list_mesh_paths_under(stage, parent_path="/World/TPV25Valve")
-    triangle_paths = [p for p in mesh_paths if p.split("/")[-1].endswith("_Triangle")]
-    vessel_path = triangle_paths[0] if triangle_paths else "/World/TPV25Valve/Mesh"
+    vessel_path = "/World/TPV25Valve/TPV25Valve_Triangle"
 else:
     vessel_path = "/World/TPV25Valve/Mesh"
 
@@ -184,8 +182,8 @@
         str(output_usd),
         vessel_path,
         color_primvar,
-        intensity_range=(colormap_range_min, colormap_range_max),
         cmap=colormap_name,
+        intensity_range=(colormap_range_min, colormap_range_max),
         use_sigmoid_scale=True,
         bind_vertex_color_material=True,
     )
diff --git a/experiments/Heart-Create_Statistical_Model/1-input_meshes_to_input_surfaces.py b/experiments/Heart-Create_Statistical_Model/1-input_meshes_to_input_surfaces.py
index 5858577..3f2c0a4 100644
--- a/experiments/Heart-Create_Statistical_Model/1-input_meshes_to_input_surfaces.py
+++ b/experiments/Heart-Create_Statistical_Model/1-input_meshes_to_input_surfaces.py
@@ -13,7 +13,7 @@
 
 import pyvista as pv
 
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 
 _HERE = Path(__file__).parent
 
@@ -77,5 +77,5 @@
     plotter = pv.Plotter()
     plotter.add_mesh(first_surface, color="lightblue", show_edges=True)
     plotter.add_axes()
-    if not running_as_test():
+    if not TestTools.running_as_test():
         plotter.show()
diff --git a/experiments/Heart-Create_Statistical_Model/2-input_surfaces_to_surfaces_aligned.py b/experiments/Heart-Create_Statistical_Model/2-input_surfaces_to_surfaces_aligned.py
index 822fe27..f0cc516 100644
--- a/experiments/Heart-Create_Statistical_Model/2-input_surfaces_to_surfaces_aligned.py
+++ b/experiments/Heart-Create_Statistical_Model/2-input_surfaces_to_surfaces_aligned.py
@@ -23,7 +23,7 @@
 import pyvista as pv
 
 from physiomotion4d.contour_tools import ContourTools
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 from physiomotion4d.register_models_icp import RegisterModelsICP
 
 _HERE = Path(__file__).parent
@@ -169,7 +169,7 @@
     plotter.show_axes()
 
     plotter.link_views()
-    if not running_as_test():
+    if not TestTools.running_as_test():
         plotter.show()
 
 # %% [markdown]
@@ -267,7 +267,7 @@
 
 plt.tight_layout()
 plt.savefig(output_dir / "registration_statistics.png", dpi=150, bbox_inches="tight")
-if not running_as_test():
+if not TestTools.running_as_test():
     plt.show()
 
 print(f"\nPlot saved to: {output_dir / 'registration_statistics.png'}")
diff --git a/experiments/Heart-Create_Statistical_Model/3-registration_based_correspondence.py b/experiments/Heart-Create_Statistical_Model/3-registration_based_correspondence.py
index 7fa0329..2baf2ef 100644
--- a/experiments/Heart-Create_Statistical_Model/3-registration_based_correspondence.py
+++ b/experiments/Heart-Create_Statistical_Model/3-registration_based_correspondence.py
@@ -28,7 +28,7 @@
 from pathlib import Path
 
 from physiomotion4d.contour_tools import ContourTools
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 from physiomotion4d.register_models_distance_maps import RegisterModelsDistanceMaps
 
 _HERE = Path(__file__).parent
@@ -222,7 +222,7 @@
 
     # Link the camera views so they rotate together
     plotter.link_views()
-    if not running_as_test():
+    if not TestTools.running_as_test():
         plotter.show()
 
 # %% [markdown]
@@ -277,7 +277,7 @@
 
     plotter.show_axes()
     plotter.camera_position = "iso"
-    if not running_as_test():
+    if not TestTools.running_as_test():
         plotter.show()
 
 # %%
@@ -321,7 +321,7 @@
     plot_file = output_dir / "registration_statistics.png"
     plt.savefig(plot_file, dpi=150, bbox_inches="tight")
     print(f"\nPlot saved to: {plot_file}")
-    if not running_as_test():
+    if not TestTools.running_as_test():
         plt.show()
 else:
     print("\nNo statistics to plot.")
diff --git a/experiments/Heart-Create_Statistical_Model/4-surfaces_aligned_correspond_to_pca_inputs.py b/experiments/Heart-Create_Statistical_Model/4-surfaces_aligned_correspond_to_pca_inputs.py
index be852dc..59eb65d 100644
--- a/experiments/Heart-Create_Statistical_Model/4-surfaces_aligned_correspond_to_pca_inputs.py
+++ b/experiments/Heart-Create_Statistical_Model/4-surfaces_aligned_correspond_to_pca_inputs.py
@@ -7,7 +7,7 @@
 import pyvista as pv
 
 from physiomotion4d.contour_tools import ContourTools
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 
 _HERE = Path(__file__).parent
 
@@ -112,7 +112,7 @@
 
     # Link the camera views so they rotate together
     plotter.link_views()
-    if not running_as_test():
+    if not TestTools.running_as_test():
         plotter.show()
 
 # %% [markdown]
@@ -167,5 +167,5 @@
 
     plotter.show_axes()
     plotter.camera_position = "iso"
-    if not running_as_test():
+    if not TestTools.running_as_test():
         plotter.show()
diff --git a/experiments/Heart-Create_Statistical_Model/5-compute_pca_model.py b/experiments/Heart-Create_Statistical_Model/5-compute_pca_model.py
index 5fb2bc3..773988d 100644
--- a/experiments/Heart-Create_Statistical_Model/5-compute_pca_model.py
+++ b/experiments/Heart-Create_Statistical_Model/5-compute_pca_model.py
@@ -22,7 +22,7 @@
 
 from sklearn.decomposition import PCA  # SparsePCA, ...
 
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 
 _HERE = Path(__file__).parent
 
@@ -243,7 +243,7 @@ def generate_pc_variation(pc_index, std_dev_multiplier=3.0):
 # Link all three views so camera movements are synchronized
 plotter.link_views()
 
-if not running_as_test():
+if not TestTools.running_as_test():
     plotter.show()
 
 # %%
@@ -267,7 +267,7 @@ def generate_pc_variation(pc_index, std_dev_multiplier=3.0):
 ax2.grid(True, alpha=0.3)
 
 plt.tight_layout()
-if not running_as_test():
+if not TestTools.running_as_test():
     plt.show()
 
 print(
diff --git a/experiments/Heart-Create_Statistical_Model/README.md b/experiments/Heart-Create_Statistical_Model/README.md
index 00b5286..b79868b 100644
--- a/experiments/Heart-Create_Statistical_Model/README.md
+++ b/experiments/Heart-Create_Statistical_Model/README.md
@@ -2,7 +2,7 @@
 
 This experiment demonstrates how to create a Principal Component Analysis (PCA) statistical shape model of a heart from a population of meshes using the KCL Heart Model dataset.
 
-**⚠️ IMPORTANT:** Users should complete this experiment BEFORE attempting the `Heart-Statistical_Model_To_Patient` experiment, as it generates the PCA model data required for patient-specific registration.
+**IMPORTANT:** Users should complete this experiment BEFORE attempting the `Heart-Statistical_Model_To_Patient` experiment, as it generates the PCA model data required for patient-specific registration.
 
 ## Overview
 
@@ -17,11 +17,11 @@ This experiment uses the **King's College London (KCL) four-chamber heart model
 
 The experiment requires the **KCL-Heart-Model** dataset located in `data/KCL-Heart-Model/`.
 
-### ⚠️ Manual Download Required
+### Manual Download Required
 
 The KCL dataset is NOT automatically downloaded. You must manually download it from:
 
-**🔗 [Virtual cohort of adult healthy four-chamber heart meshes from CT images](https://zenodo.org/records/4590294)**
+**[Virtual cohort of adult healthy four-chamber heart meshes from CT images](https://zenodo.org/records/4590294)**
 
 See `data/KCL-Heart-Model/README.md` for complete download and setup instructions.
 
@@ -33,33 +33,35 @@ See `data/KCL-Heart-Model/README.md` for complete download and setup instruction
 
 ## Pipeline Overview
 
-This experiment follows a fully automated multi-step process using Jupyter notebooks:
+This experiment follows a fully automated multi-step process. Each step is a
+`# %%` percent-cell Python script that can be run top-to-bottom with
+`python <script>.py` or stepped through cell-by-cell in VS Code / Cursor.
 
-### Automated Steps (All Jupyter Notebooks)
+### Automated Steps
 
-1. **`1-input_meshes_to_input_surfaces.ipynb`**
+1. **`1-input_meshes_to_input_surfaces.py`**
    - Converts volumetric tetrahedral meshes to surface representations
    - Extracts relevant anatomical surfaces for shape analysis
    - Prepares data for alignment
 
-2. **`2-input_surfaces_to_surfaces_aligned.ipynb`**
+2. **`2-input_surfaces_to_surfaces_aligned.py`**
    - Performs initial rigid alignment of surfaces using ICP
    - Centers and orients meshes for consistent positioning
    - Computes and saves the average surface from aligned meshes
    - Prepares aligned data for correspondence computation
 
-3. **`3-registration_based_correspondence.ipynb`**
+3. **`3-registration_based_correspondence.py`**
    - Establishes point correspondences across the population using ANTs SyN deformable registration
    - Uses mask-based distance map registration via `RegisterModelsDistanceMaps`
    - Performs diffeomorphic (smooth, invertible) deformation to the average surface
    - Critical step for meaningful PCA analysis
 
-4. **`4-surfaces_aligned_correspond_to_pca_inputs.ipynb`**
+4. **`4-surfaces_aligned_correspond_to_pca_inputs.py`**
    - Converts corresponded surfaces to PCA input format
    - Prepares data matrices for statistical analysis
    - Validates correspondence quality
 
-5. **`5-compute_pca_model.ipynb`**
+5. **`5-compute_pca_model.py`**
    - Computes PCA on the corresponded point sets using sklearn
    - Generates eigenvectors, eigenvalues, and explained variance
    - Exports complete PCA model to `pca_model.json`
@@ -97,19 +99,18 @@ Statistical shape modeling uses **sklearn's PCA implementation**:
 
 See `data/KCL-Heart-Model/README.md` for download instructions. Place downloaded mesh files in `data/KCL-Heart-Model/input_meshes/`.
 
-### 2. Run All Notebooks in Sequence
+### 2. Run All Scripts in Sequence
 
 ```bash
-# Open Jupyter Lab in the experiment directory
 cd experiments/Heart-Create_Statistical_Model/
-jupyter lab
-
-# Execute notebooks in order:
-# 1. 1-input_meshes_to_input_surfaces.ipynb     - Extract surfaces from volumetric meshes
-# 2. 2-input_surfaces_to_surfaces_aligned.ipynb - Rigid ICP alignment + compute average
-# 3. 3-registration_based_correspondence.ipynb  - ANTs SyN deformable correspondence
-# 4. 4-surfaces_aligned_correspond_to_pca_inputs.ipynb - Prepare PCA input matrices
-# 5. 5-compute_pca_model.ipynb                  - Compute PCA and export JSON model
+
+# Execute scripts in order (each can also be stepped through cell-by-cell
+# in VS Code or Cursor via the `# %%` cell markers):
+python 1-input_meshes_to_input_surfaces.py     # Extract surfaces from volumetric meshes
+python 2-input_surfaces_to_surfaces_aligned.py # Rigid ICP alignment + compute average
+python 3-registration_based_correspondence.py  # ANTs SyN deformable correspondence
+python 4-surfaces_aligned_correspond_to_pca_inputs.py  # Prepare PCA input matrices
+python 5-compute_pca_model.py                  # Compute PCA and export JSON model
 ```
 
 **Total Runtime:** Approximately 2-4 hours depending on hardware (20 heart meshes, ANTs registration is computationally intensive).
@@ -134,7 +135,7 @@ After completing this experiment, you will have generated files in `kcl-heart-mo
 - **`pca_inputs/`** - Point coordinate matrices ready for PCA computation
 
 ### Visualizations
-- Various visualization plots generated within notebooks showing:
+- Various visualization plots generated within the scripts showing:
   - Alignment quality
   - Deformation magnitude maps
   - PCA mode variations
@@ -162,8 +163,9 @@ registered_mesh = workflow.run_workflow()
 ## Requirements
 
 ### Software
-- Python 3.10+ with PhysioMotion4D installed
-- Jupyter Lab or Jupyter Notebook
+- Python 3.11+ with PhysioMotion4D installed
+- VS Code or Cursor with the Python extension for cell-by-cell execution
+  (optional; scripts also run end-to-end as plain Python)
 - ITK, VTK, PyVista (included with PhysioMotion4D)
 - ANTs (Advanced Normalization Tools) - installed automatically with PhysioMotion4D
 - scikit-learn for PCA computation
diff --git a/experiments/Heart-GatedCT_To_USD/0-download_and_convert_4d_to_3d.py b/experiments/Heart-GatedCT_To_USD/0-download_and_convert_4d_to_3d.py
index 83ef89e..eaada36 100644
--- a/experiments/Heart-GatedCT_To_USD/0-download_and_convert_4d_to_3d.py
+++ b/experiments/Heart-GatedCT_To_USD/0-download_and_convert_4d_to_3d.py
@@ -1,34 +1,26 @@
 #!/usr/bin/env python
 # %%
-import os
 import shutil
-import urllib.request
+from pathlib import Path
 
 from physiomotion4d.convert_nrrd_4d_to_3d import ConvertNRRD4DTo3D
+from physiomotion4d.data_download_tools import DataDownloadTools
 
-_HERE = os.path.dirname(os.path.abspath(__file__))
+_HERE = Path(__file__).resolve().parent
 
 # %%
-data_dir = os.path.join(_HERE, "..", "..", "data", "Slicer-Heart-CT")
-output_dir = os.path.join(_HERE, "results")
+data_dir = _HERE.parent.parent / "data" / "Slicer-Heart-CT"
+output_dir = _HERE / "results"
 
-if not os.path.exists(data_dir):
-    os.makedirs(data_dir)
+data_dir.mkdir(parents=True, exist_ok=True)
+output_dir.mkdir(parents=True, exist_ok=True)
 
-if not os.path.exists(output_dir):
-    os.makedirs(output_dir)
-
-# %%
-input_image_url = "https://github.com/SlicerHeart/SlicerHeart/releases/download/TestingData/TruncalValve_4DCT.seq.nrrd"
-input_image_filename = os.path.join(data_dir, "TruncalValve_4DCT.seq.nrrd")
-
-if not os.path.exists(input_image_filename):
-    urllib.request.urlretrieve(input_image_url, input_image_filename)
+input_image_filename = DataDownloadTools.DownloadSlicerHeartCTData(data_dir)
 
 # %%
 conv = ConvertNRRD4DTo3D()
-conv.load_nrrd_4d(f"{data_dir}/TruncalValve_4DCT.seq.nrrd")
-conv.save_3d_images(f"{data_dir}/slice")
+conv.load_nrrd_4d(str(input_image_filename))
+conv.save_3d_images(data_dir, "slice")
 
 # Save the mid-stroke slice as the fixed/reference image
-shutil.copyfile(f"{data_dir}/slice_007.mha", f"{output_dir}/slice_fixed.mha")
+shutil.copyfile(data_dir / "slice_007.mha", output_dir / "slice_fixed.mha")
diff --git a/experiments/Heart-GatedCT_To_USD/1-register_images.py b/experiments/Heart-GatedCT_To_USD/1-register_images.py
index 75f45ec..2ac45cf 100644
--- a/experiments/Heart-GatedCT_To_USD/1-register_images.py
+++ b/experiments/Heart-GatedCT_To_USD/1-register_images.py
@@ -1,84 +1,41 @@
 #!/usr/bin/env python
 # %%
-import os
+from pathlib import Path
 
 import itk
 
 from physiomotion4d.register_images_ants import RegisterImagesANTs
 from physiomotion4d.segment_chest_total_segmentator import SegmentChestTotalSegmentator
+from physiomotion4d.test_tools import TestTools
 from physiomotion4d.transform_tools import TransformTools
 
-_HERE = os.path.dirname(os.path.abspath(__file__))
+# nnUNetv2 (used by TotalSegmentator) spawns a multiprocessing.Pool. On Windows
+# the spawn start method re-imports this script in each child; without the
+# __name__ == "__main__" guard around the top-level work, that re-import fires
+# segment() again and Python's spawn-cascade detector raises RuntimeError.
+if __name__ == "__main__":
+    test_mode = TestTools.running_as_test()
 
-# %%
-# Number of cardiac frames and step size; downstream scripts must use the same values.
-N_FRAMES = 21
-FRAME_STEP = (
-    4  # Set to 1 to process all frames; 4 to process every 4th (faster testing)
-)
-
-data_dir = os.path.join(_HERE, "..", "..", "data", "Slicer-Heart-CT")
+    _HERE = Path(__file__).resolve().parent
 
-output_dir = os.path.join(_HERE, "results")
-if not os.path.exists(output_dir):
-    os.makedirs(output_dir)
+    # %%
+    # Number of cardiac frames and step size; downstream scripts must use the same values.
+    N_FRAMES = 21
+    FRAME_STEP = 4 if test_mode else 1
 
-fixed_image_filename = os.path.join(output_dir, "slice_fixed.mha")
-fixed_image = itk.imread(fixed_image_filename)
+    data_dir = _HERE.parent.parent / "data" / "Slicer-Heart-CT"
 
-# %%
-seg = SegmentChestTotalSegmentator()
-seg.contrast_threshold = 500
-result = seg.segment(fixed_image, contrast_enhanced_study=True)
-labelmap_mask = result["labelmap"]
-lung_mask = result["lung"]
-heart_mask = result["heart"]
-major_vessels_mask = result["major_vessels"]
-bone_mask = result["bone"]
-other_mask = result["other"]
-contrast_mask = result["contrast"]
-
-fixed_image_labelmap = labelmap_mask
-itk.imwrite(
-    fixed_image_labelmap,
-    os.path.join(output_dir, "slice_fixed_mask.mha"),
-    compression=True,
-)
-
-heart_arr = itk.GetArrayFromImage(heart_mask)
-contrast_arr = itk.GetArrayFromImage(contrast_mask)
-major_vessels_arr = itk.GetArrayFromImage(major_vessels_mask)
-fixed_image_dynamic_anatomy_mask = itk.GetImageFromArray(
-    heart_arr + contrast_arr + major_vessels_arr
-)
-fixed_image_dynamic_anatomy_mask.CopyInformation(fixed_image)
-itk.imwrite(
-    fixed_image_dynamic_anatomy_mask,
-    os.path.join(output_dir, "slice_fixed.dynamic_anatomy_mask.mha"),
-    compression=True,
-)
-
-lung_arr = itk.GetArrayFromImage(lung_mask)
-bone_arr = itk.GetArrayFromImage(bone_mask)
-other_arr = itk.GetArrayFromImage(other_mask)
-fixed_image_static_mask = itk.GetImageFromArray(lung_arr + bone_arr + other_arr)
-fixed_image_static_mask.CopyInformation(fixed_image)
-itk.imwrite(
-    fixed_image_static_mask,
-    os.path.join(output_dir, "slice_fixed.static_anatomy_mask.mha"),
-    compression=True,
-)
+    output_dir = _HERE / "results"
+    output_dir.mkdir(parents=True, exist_ok=True)
 
-# %%
-reg = RegisterImagesANTs()
-reg.set_mask_dilation(5)
-reg.set_number_of_iterations([10, 5, 2])
+    fixed_image_filename = output_dir / "slice_fixed.mha"
+    fixed_image = itk.imread(str(fixed_image_filename))
 
-# %%
-for i in range(0, N_FRAMES, FRAME_STEP):
-    print(f"Processing slice {i:03d}")
-    moving_image = itk.imread(os.path.join(data_dir, f"slice_{i:03d}.mha"))
-    result = seg.segment(moving_image, contrast_enhanced_study=True)
+    # %%
+    seg = SegmentChestTotalSegmentator()
+    seg.contrast_threshold = 500
+    result = seg.segment(fixed_image, contrast_enhanced_study=True)
+    # %%
     labelmap_mask = result["labelmap"]
     lung_mask = result["lung"]
     heart_mask = result["heart"]
@@ -86,103 +43,155 @@
     bone_mask = result["bone"]
     other_mask = result["other"]
     contrast_mask = result["contrast"]
-    itk.imwrite(
-        labelmap_mask,
-        os.path.join(output_dir, f"slice_{i:03d}_mask.mha"),
-        compression=True,
-    )
 
-    # Register the whole image
-    reg.set_fixed_image(fixed_image)
-    results = reg.register(moving_image)
-    inverse_transform = results["inverse_transform"]
-    forward_transform = results["forward_transform"]
-    moving_image_reg = TransformTools().transform_image(
-        moving_image, forward_transform, fixed_image, "sinc"
-    )  # Final resampling with sinc
+    fixed_image_labelmap = labelmap_mask
     itk.imwrite(
-        moving_image_reg,
-        os.path.join(output_dir, f"slice_{i:03d}.reg_all.mha"),
-        compression=True,
-    )
-    itk.transformwrite(
-        [forward_transform],
-        os.path.join(output_dir, f"slice_{i:03d}.reg_all.forward.hdf"),
-        compression=True,
-    )
-    itk.transformwrite(
-        [inverse_transform],
-        os.path.join(output_dir, f"slice_{i:03d}.reg_all.inverse.hdf"),
+        fixed_image_labelmap,
+        str(output_dir / "slice_fixed_mask.mha"),
         compression=True,
     )
 
-    # Register the dynamic anatomy mask
     heart_arr = itk.GetArrayFromImage(heart_mask)
     contrast_arr = itk.GetArrayFromImage(contrast_mask)
     major_vessels_arr = itk.GetArrayFromImage(major_vessels_mask)
-    dynamic_anatomy_arr = heart_arr + contrast_arr + major_vessels_arr
-    moving_image_dynamic_anatomy_mask = itk.GetImageFromArray(dynamic_anatomy_arr)
-    moving_image_dynamic_anatomy_mask.CopyInformation(moving_image)
-    reg.set_fixed_image(fixed_image)
-    reg.set_fixed_mask(fixed_image_dynamic_anatomy_mask)
-    results = reg.register(moving_image, moving_image_dynamic_anatomy_mask)
-    inverse_transform = results["inverse_transform"]
-    forward_transform = results["forward_transform"]
-    moving_image_reg_dynamic_anatomy = TransformTools().transform_image(
-        moving_image, forward_transform, fixed_image, "sinc"
-    )  # Final resampling with sinc
-    itk.imwrite(
-        moving_image_dynamic_anatomy_mask,
-        os.path.join(output_dir, f"slice_{i:03d}.dynamic_anatomy_mask.mha"),
-        compression=True,
+    fixed_image_dynamic_anatomy_mask = itk.GetImageFromArray(
+        heart_arr + contrast_arr + major_vessels_arr
     )
+    fixed_image_dynamic_anatomy_mask.CopyInformation(fixed_image)
     itk.imwrite(
-        moving_image_reg_dynamic_anatomy,
-        os.path.join(output_dir, f"slice_{i:03d}.reg_dynamic_anatomy.mha"),
-        compression=True,
-    )
-    itk.transformwrite(
-        [forward_transform],
-        os.path.join(output_dir, f"slice_{i:03d}.reg_dynamic_anatomy.forward.hdf"),
-        compression=True,
-    )
-    itk.transformwrite(
-        [inverse_transform],
-        os.path.join(output_dir, f"slice_{i:03d}.reg_dynamic_anatomy.inverse.hdf"),
+        fixed_image_dynamic_anatomy_mask,
+        str(output_dir / "slice_fixed.dynamic_anatomy_mask.mha"),
         compression=True,
     )
 
-    # Register the static anatomy mask
     lung_arr = itk.GetArrayFromImage(lung_mask)
     bone_arr = itk.GetArrayFromImage(bone_mask)
     other_arr = itk.GetArrayFromImage(other_mask)
-    moving_image_static_mask = itk.GetImageFromArray(lung_arr + bone_arr + other_arr)
-    moving_image_static_mask.CopyInformation(moving_image)
-    reg.set_fixed_image(fixed_image)
-    reg.set_fixed_mask(fixed_image_static_mask)
-    results = reg.register(moving_image, moving_image_static_mask)
-    inverse_transform = results["inverse_transform"]
-    forward_transform = results["forward_transform"]
-    moving_image_reg_static = TransformTools().transform_image(
-        moving_image, forward_transform, fixed_image, "sinc"
-    )  # Final resampling with sinc
+    fixed_image_static_mask = itk.GetImageFromArray(lung_arr + bone_arr + other_arr)
+    fixed_image_static_mask.CopyInformation(fixed_image)
     itk.imwrite(
-        moving_image_static_mask,
-        os.path.join(output_dir, f"slice_{i:03d}.static_anatomy_mask.mha"),
-        compression=True,
-    )
-    itk.imwrite(
-        moving_image_reg_static,
-        os.path.join(output_dir, f"slice_{i:03d}.reg_static_anatomy.mha"),
-        compression=True,
-    )
-    itk.transformwrite(
-        [forward_transform],
-        os.path.join(output_dir, f"slice_{i:03d}.reg_static_anatomy.forward.hdf"),
-        compression=True,
-    )
-    itk.transformwrite(
-        [inverse_transform],
-        os.path.join(output_dir, f"slice_{i:03d}.reg_static_anatomy.inverse.hdf"),
+        fixed_image_static_mask,
+        str(output_dir / "slice_fixed.static_anatomy_mask.mha"),
         compression=True,
     )
+
+    # %%
+    reg = RegisterImagesANTs()
+    reg.set_mask_dilation(5)
+    reg.set_number_of_iterations([10, 5, 2])
+
+    # %%
+    for i in range(0, N_FRAMES, FRAME_STEP):
+        print(f"Processing slice {i:03d}")
+        moving_image = itk.imread(str(data_dir / f"slice_{i:03d}.mha"))
+        result = seg.segment(moving_image, contrast_enhanced_study=True)
+        labelmap_mask = result["labelmap"]
+        lung_mask = result["lung"]
+        heart_mask = result["heart"]
+        major_vessels_mask = result["major_vessels"]
+        bone_mask = result["bone"]
+        other_mask = result["other"]
+        contrast_mask = result["contrast"]
+        itk.imwrite(
+            labelmap_mask,
+            str(output_dir / f"slice_{i:03d}_mask.mha"),
+            compression=True,
+        )
+
+        # Register the whole image
+        reg.set_fixed_image(fixed_image)
+        reg.set_fixed_mask(None)
+        results = reg.register(moving_image)
+        inverse_transform = results["inverse_transform"]
+        forward_transform = results["forward_transform"]
+        moving_image_reg = TransformTools().transform_image(
+            moving_image, forward_transform, fixed_image, "sinc"
+        )  # Final resampling with sinc
+        itk.imwrite(
+            moving_image_reg,
+            str(output_dir / f"slice_{i:03d}.reg_all.mha"),
+            compression=True,
+        )
+        itk.transformwrite(
+            [forward_transform],
+            str(output_dir / f"slice_{i:03d}.reg_all.forward.hdf"),
+            compression=True,
+        )
+        itk.transformwrite(
+            [inverse_transform],
+            str(output_dir / f"slice_{i:03d}.reg_all.inverse.hdf"),
+            compression=True,
+        )
+
+        # Register the dynamic anatomy mask
+        heart_arr = itk.GetArrayFromImage(heart_mask)
+        contrast_arr = itk.GetArrayFromImage(contrast_mask)
+        major_vessels_arr = itk.GetArrayFromImage(major_vessels_mask)
+        dynamic_anatomy_arr = heart_arr + contrast_arr + major_vessels_arr
+        moving_image_dynamic_anatomy_mask = itk.GetImageFromArray(dynamic_anatomy_arr)
+        moving_image_dynamic_anatomy_mask.CopyInformation(moving_image)
+        reg.set_fixed_image(fixed_image)
+        reg.set_fixed_mask(fixed_image_dynamic_anatomy_mask)
+        results = reg.register(moving_image, moving_image_dynamic_anatomy_mask)
+        inverse_transform = results["inverse_transform"]
+        forward_transform = results["forward_transform"]
+        moving_image_reg_dynamic_anatomy = TransformTools().transform_image(
+            moving_image, forward_transform, fixed_image, "sinc"
+        )  # Final resampling with sinc
+        itk.imwrite(
+            moving_image_dynamic_anatomy_mask,
+            str(output_dir / f"slice_{i:03d}.dynamic_anatomy_mask.mha"),
+            compression=True,
+        )
+        itk.imwrite(
+            moving_image_reg_dynamic_anatomy,
+            str(output_dir / f"slice_{i:03d}.reg_dynamic_anatomy.mha"),
+            compression=True,
+        )
+        itk.transformwrite(
+            [forward_transform],
+            str(output_dir / f"slice_{i:03d}.reg_dynamic_anatomy.forward.hdf"),
+            compression=True,
+        )
+        itk.transformwrite(
+            [inverse_transform],
+            str(output_dir / f"slice_{i:03d}.reg_dynamic_anatomy.inverse.hdf"),
+            compression=True,
+        )
+
+        # Register the static anatomy mask
+        lung_arr = itk.GetArrayFromImage(lung_mask)
+        bone_arr = itk.GetArrayFromImage(bone_mask)
+        other_arr = itk.GetArrayFromImage(other_mask)
+        moving_image_static_mask = itk.GetImageFromArray(
+            lung_arr + bone_arr + other_arr
+        )
+        moving_image_static_mask.CopyInformation(moving_image)
+        reg.set_fixed_image(fixed_image)
+        reg.set_fixed_mask(fixed_image_static_mask)
+        results = reg.register(moving_image, moving_image_static_mask)
+        inverse_transform = results["inverse_transform"]
+        forward_transform = results["forward_transform"]
+        moving_image_reg_static = TransformTools().transform_image(
+            moving_image, forward_transform, fixed_image, "sinc"
+        )  # Final resampling with sinc
+        itk.imwrite(
+            moving_image_static_mask,
+            str(output_dir / f"slice_{i:03d}.static_anatomy_mask.mha"),
+            compression=True,
+        )
+        itk.imwrite(
+            moving_image_reg_static,
+            str(output_dir / f"slice_{i:03d}.reg_static_anatomy.mha"),
+            compression=True,
+        )
+        itk.transformwrite(
+            [forward_transform],
+            str(output_dir / f"slice_{i:03d}.reg_static_anatomy.forward.hdf"),
+            compression=True,
+        )
+        itk.transformwrite(
+            [inverse_transform],
+            str(output_dir / f"slice_{i:03d}.reg_static_anatomy.inverse.hdf"),
+            compression=True,
+        )
diff --git a/experiments/Heart-GatedCT_To_USD/2-generate_segmentation.py b/experiments/Heart-GatedCT_To_USD/2-generate_segmentation.py
index 636ae11..6678984 100644
--- a/experiments/Heart-GatedCT_To_USD/2-generate_segmentation.py
+++ b/experiments/Heart-GatedCT_To_USD/2-generate_segmentation.py
@@ -7,152 +7,159 @@
 import pyvista as pv
 
 from physiomotion4d.contour_tools import ContourTools
-from physiomotion4d.notebook_utils import running_as_test
 from physiomotion4d.segment_chest_total_segmentator import SegmentChestTotalSegmentator
+from physiomotion4d.test_tools import TestTools
+
+# nnUNetv2 (used by TotalSegmentator) spawns a multiprocessing.Pool. On Windows
+# the spawn start method re-imports this script in each child; without the
+# __name__ == "__main__" guard around the top-level work, that re-import fires
+# segment() again and Python's spawn-cascade detector raises RuntimeError.
+if __name__ == "__main__":
+    _HERE = os.path.dirname(os.path.abspath(__file__))
+
+    # %%
+    # When re-running, you can bypass certain long-running steps
+    re_run_image_max = True
+    use_fixed_image = True
+    re_run_image_segmentation = True
+
+    # %%
+    output_dir = os.path.join(_HERE, "results")
+    max_image = None
+    print("Computing max image...")
+    if re_run_image_max and not use_fixed_image:
+        # Compute max of all images
+        image = None
+        try:
+            image = itk.imread(
+                os.path.join(output_dir, "slice_000.reg_dynamic_anatomy.mha")
+            )
+        except (FileNotFoundError, OSError):
+            print("No image found. Aborting. Please run 1-generate_images.ipynb first.")
+            exit(1)
+        arr = itk.array_from_image(image)
+        print(arr.shape)
+        arr = np.where(arr == 0, -1000, arr)
+        for i in range(0, 21):
+            print(f"Processing slice {i:03d}...")
+            tmp_arr = itk.array_from_image(
+                itk.imread(
+                    os.path.join(output_dir, f"slice_{i:03d}.reg_dynamic_anatomy.mha")
+                )
+            )
+            tmp_arr = np.where(tmp_arr == 0, -1000, tmp_arr)
+            arr = np.maximum(arr, tmp_arr)
+        print("Max image computed.")
+        max_image = itk.image_from_array(arr)
+        max_image.CopyInformation(image)
+        itk.imwrite(
+            max_image,
+            os.path.join(output_dir, "slice_max.reg_dynamic_anatomy.mha"),
+            compression=True,
+        )
 
-_HERE = os.path.dirname(os.path.abspath(__file__))
-
-# %%
-# When re-running, you can bypass certain long-running steps
-re_run_image_max = True
-use_fixed_image = True
-re_run_image_segmentation = True
-
-# %%
-output_dir = os.path.join(_HERE, "results")
-max_image = None
-print("Computing max image...")
-if re_run_image_max and not use_fixed_image:
-    # Compute max of all images
-    image = None
-    try:
-        image = itk.imread(
-            os.path.join(output_dir, "slice_000.reg_dynamic_anatomy.mha")
+    # %%
+    if use_fixed_image:
+        max_image = itk.imread(os.path.join(output_dir, "slice_fixed.mha"))
+        outname = "slice_fixed"
+    else:
+        max_image = itk.imread(
+            os.path.join(output_dir, "slice_max.reg_dynamic_anatomy.mha")
         )
-    except (FileNotFoundError, OSError):
-        print("No image found. Aborting. Please run 1-generate_images.ipynb first.")
-        exit(1)
-    arr = itk.array_from_image(image)
-    print(arr.shape)
-    arr = np.where(arr == 0, -1000, arr)
-    for i in range(4, 21, 4):  # Process every 4th to save time for testing
-        print(f"Processing slice {i:03d}...")
-        tmp_arr = itk.array_from_image(
-            itk.imread(
-                os.path.join(output_dir, f"slice_{i:03d}.reg_dynamic_anatomy.mha")
-            )
+        outname = "slice_max"
+
+    seg = SegmentChestTotalSegmentator()
+    seg.contrast_threshold = 500
+    if re_run_image_segmentation:
+        result = seg.segment(max_image, contrast_enhanced_study=True)
+        labelmap_image = result["labelmap"]
+        itk.imwrite(
+            labelmap_image,
+            os.path.join(output_dir, f"{outname}.all_mask.mha"),
+            compression=True,
         )
-        tmp_arr = np.where(tmp_arr == 0, -1000, tmp_arr)
-        arr = np.maximum(arr, tmp_arr)
-    print("Max image computed.")
-    max_image = itk.image_from_array(arr)
-    max_image.CopyInformation(image)
+    else:
+        labelmap_image = itk.imread(os.path.join(output_dir, f"{outname}.all_mask.mha"))
+        result = seg.create_anatomy_group_masks(labelmap_image)
+
+    lung_mask = result["lung"]
+    heart_mask = result["heart"]
+    major_vessels_mask = result["major_vessels"]
+    bone_mask = result["bone"]
+    soft_tissue_mask = result["soft_tissue"]
+    other_mask = result["other"]
+    contrast_mask = result["contrast"]
+
+    # %%
+    con = ContourTools()
+    all_contours = con.extract_contours(labelmap_image)
+    all_contours.save(os.path.join(output_dir, f"{outname}.all_mask.vtp"))
+
+    # %%
+    label_arr = itk.array_from_image(labelmap_image)
+    lung_arr = itk.array_from_image(lung_mask)
+    heart_arr = itk.array_from_image(heart_mask)
+    major_vessels_arr = itk.array_from_image(major_vessels_mask)
+    bone_arr = itk.array_from_image(bone_mask)
+    soft_tissue_arr = itk.array_from_image(soft_tissue_mask)
+    other_arr = itk.array_from_image(other_mask)
+    contrast_arr = itk.array_from_image(contrast_mask)
+
+    # %%
+    dynamic_anatomy_arr = np.maximum(heart_arr, contrast_arr)
+    dynamic_anatomy_arr = np.maximum(dynamic_anatomy_arr, major_vessels_arr)
+    dynamic_anatomy_arr = np.where(dynamic_anatomy_arr, label_arr, 0)
+    dynamic_anatomy_image = itk.image_from_array(dynamic_anatomy_arr.astype(np.int16))
+    dynamic_anatomy_image.CopyInformation(labelmap_image)
     itk.imwrite(
-        max_image,
-        os.path.join(output_dir, "slice_max.reg_dynamic_anatomy.mha"),
+        dynamic_anatomy_image,
+        os.path.join(output_dir, f"{outname}.dynamic_anatomy_mask.mha"),
         compression=True,
     )
 
-# %%
-if use_fixed_image:
-    max_image = itk.imread(os.path.join(output_dir, "slice_fixed.mha"))
-    outname = "slice_fixed"
-else:
-    max_image = itk.imread(
-        os.path.join(output_dir, "slice_max.reg_dynamic_anatomy.mha")
-    )
-    outname = "slice_max"
+    contours = con.extract_contours(dynamic_anatomy_image)
+    contours.save(os.path.join(output_dir, f"{outname}.dynamic_anatomy_mask.vtp"))
 
-seg = SegmentChestTotalSegmentator()
-seg.contrast_threshold = 500
-if re_run_image_segmentation:
-    result = seg.segment(max_image, contrast_enhanced_study=True)
-    labelmap_image = result["labelmap"]
+    # %%
+    static_anatomy_arr = lung_arr + bone_arr + soft_tissue_arr + other_arr
+    static_anatomy_arr = np.where(static_anatomy_arr, label_arr, 0)
+    static_anatomy_image = itk.image_from_array(static_anatomy_arr.astype(np.int16))
+    static_anatomy_image.CopyInformation(labelmap_image)
     itk.imwrite(
-        labelmap_image,
-        os.path.join(output_dir, f"{outname}.all_mask.mha"),
+        static_anatomy_image,
+        os.path.join(output_dir, f"{outname}.static_anatomy_mask.mha"),
         compression=True,
     )
-else:
-    labelmap_image = itk.imread(os.path.join(output_dir, f"{outname}.all_mask.mha"))
-    result = seg.create_anatomy_group_masks(labelmap_image)
-
-lung_mask = result["lung"]
-heart_mask = result["heart"]
-major_vessels_mask = result["major_vessels"]
-bone_mask = result["bone"]
-soft_tissue_mask = result["soft_tissue"]
-other_mask = result["other"]
-contrast_mask = result["contrast"]
 
-# %%
-con = ContourTools()
-all_contours = con.extract_contours(labelmap_image)
-all_contours.save(os.path.join(output_dir, f"{outname}.all_mask.vtp"))
+    contours = con.extract_contours(static_anatomy_image)
+    contours.save(os.path.join(output_dir, f"{outname}.static_anatomy_mask.vtp"))
 
-# %%
-label_arr = itk.array_from_image(labelmap_image)
-lung_arr = itk.array_from_image(lung_mask)
-heart_arr = itk.array_from_image(heart_mask)
-major_vessels_arr = itk.array_from_image(major_vessels_mask)
-bone_arr = itk.array_from_image(bone_mask)
-soft_tissue_arr = itk.array_from_image(soft_tissue_mask)
-other_arr = itk.array_from_image(other_mask)
-contrast_arr = itk.array_from_image(contrast_mask)
-
-# %%
-dynamic_anatomy_arr = np.maximum(heart_arr, contrast_arr)
-dynamic_anatomy_arr = np.maximum(dynamic_anatomy_arr, major_vessels_arr)
-dynamic_anatomy_arr = np.where(dynamic_anatomy_arr, label_arr, 0)
-dynamic_anatomy_image = itk.image_from_array(dynamic_anatomy_arr.astype(np.int16))
-dynamic_anatomy_image.CopyInformation(labelmap_image)
-itk.imwrite(
-    dynamic_anatomy_image,
-    os.path.join(output_dir, f"{outname}.dynamic_anatomy_mask.mha"),
-    compression=True,
-)
-
-contours = con.extract_contours(dynamic_anatomy_image)
-contours.save(os.path.join(output_dir, f"{outname}.dynamic_anatomy_mask.vtp"))
+    # %%
+    input_image = None
+    if use_fixed_image:
+        input_image = itk.imread(os.path.join(output_dir, "slice_fixed.mha"), itk.SS)
+    else:
+        input_image = itk.imread(
+            os.path.join(output_dir, "slice_max.reg_dynamic_anatomy.mha"), itk.SS
+        )
+    arr = itk.array_from_image(input_image)
+    flipped_input_image = itk.image_from_array(arr)
+    flipped_input_image.CopyInformation(input_image)
 
-# %%
-static_anatomy_arr = lung_arr + bone_arr + soft_tissue_arr + other_arr
-static_anatomy_arr = np.where(static_anatomy_arr, label_arr, 0)
-static_anatomy_image = itk.image_from_array(static_anatomy_arr.astype(np.int16))
-static_anatomy_image.CopyInformation(labelmap_image)
-itk.imwrite(
-    static_anatomy_image,
-    os.path.join(output_dir, f"{outname}.static_anatomy_mask.mha"),
-    compression=True,
-)
-
-contours = con.extract_contours(static_anatomy_image)
-contours.save(os.path.join(output_dir, f"{outname}.static_anatomy_mask.vtp"))
+    image = pv.wrap(itk.vtk_image_from_image(flipped_input_image))
 
-# %%
-input_image = None
-if use_fixed_image:
-    input_image = itk.imread(os.path.join(output_dir, "slice_fixed.mha"), itk.SS)
-else:
-    input_image = itk.imread(
-        os.path.join(output_dir, "slice_max.reg_dynamic_anatomy.mha"), itk.SS
+    pl = pv.Plotter()
+    pl.add_mesh(
+        image.slice(normal="z"), cmap="bone", show_scalar_bar=False, opacity=0.5
+    )
+    pl.add_mesh(
+        contours.slice(normal="z"),
+        cmap="pink",
+        clim=[50, 800],
+        show_scalar_bar=False,
+        opacity=1.0,
     )
-arr = itk.array_from_image(input_image)
-flipped_input_image = itk.image_from_array(arr)
-flipped_input_image.CopyInformation(input_image)
-
-image = pv.wrap(itk.vtk_image_from_image(flipped_input_image))
-
-pl = pv.Plotter()
-pl.add_mesh(image.slice(normal="z"), cmap="bone", show_scalar_bar=False, opacity=0.5)
-pl.add_mesh(
-    contours.slice(normal="z"),
-    cmap="pink",
-    clim=[50, 800],
-    show_scalar_bar=False,
-    opacity=1.0,
-)
-pl.set_background("black")
-pl.camera_position = "xy"
-if not running_as_test():
-    pl.show()
+    pl.set_background("black")
+    pl.camera_position = "xy"
+    if not TestTools.running_as_test():
+        pl.show()
diff --git a/experiments/Heart-GatedCT_To_USD/3-transform_dynamic_and_static_contours.py b/experiments/Heart-GatedCT_To_USD/3-transform_dynamic_and_static_contours.py
index 848277b..5bad505 100644
--- a/experiments/Heart-GatedCT_To_USD/3-transform_dynamic_and_static_contours.py
+++ b/experiments/Heart-GatedCT_To_USD/3-transform_dynamic_and_static_contours.py
@@ -5,123 +5,135 @@
 import itk
 import pyvista as pv
 
-from physiomotion4d.contour_tools import ContourTools
 from physiomotion4d import ConvertVTKToUSD
+from physiomotion4d.contour_tools import ContourTools
 from physiomotion4d.segment_chest_total_segmentator import SegmentChestTotalSegmentator
+from physiomotion4d.test_tools import TestTools
 from physiomotion4d.usd_anatomy_tools import USDAnatomyTools
 
-_HERE = os.path.dirname(os.path.abspath(__file__))
+# Defensive: this script only reads `seg.all_mask_ids` today, but if anyone
+# ever adds a `seg.segment(...)` call it would trigger the nnUNet
+# multiprocessing.Pool which re-imports the script on Windows (spawn start
+# method) and crashes with a spawn-cascade RuntimeError. Guard pre-emptively.
+if __name__ == "__main__":
+    test_mode = TestTools.running_as_test()
+    _HERE = os.path.dirname(os.path.abspath(__file__))
 
-# Must match N_FRAMES / FRAME_STEP in 1-register_images.py
-N_FRAMES = 21
-FRAME_STEP = 4
+    # Must match N_FRAMES / FRAME_STEP in 1-register_images.py
+    N_FRAMES = 21
+    FRAME_STEP = 4 if test_mode else 1
 
-# %%
-output_dir = os.path.join(_HERE, "results")
+    # %%
+    output_dir = os.path.join(_HERE, "results")
 
-base_name = "slice_fixed"
-# base_name = "slice_max.reg_dynamic_anatomy"
+    base_name = "slice_fixed"
+    # base_name = "slice_max.reg_dynamic_anatomy"
 
-project_name = "Slicer_CardiacGatedCT"
+    project_name = "Slicer_CardiacGatedCT"
 
-do_transform_contours = True
+    do_transform_contours = True
 
+    # %%
+    def transform_contours(
+        contours, transform_filenames, frame_indices, base_name, output_dir
+    ):
+        con = ContourTools()
+        for i, transform_filename in zip(frame_indices, transform_filenames):
+            forward_transform = itk.transformread(transform_filename)[0]
+            print(f"Applying transform {transform_filename} to {base_name}")
 
-# %%
-def transform_contours(
-    contours, transform_filenames, frame_indices, base_name, output_dir
-):
-    con = ContourTools()
-    for i, transform_filename in zip(frame_indices, transform_filenames):
-        forward_transform = itk.transformread(transform_filename)[0]
-        print(f"Applying transform {transform_filename} to {base_name}")
-
-        new_contours = con.transform_contours(
-            contours, forward_transform, with_deformation_magnitude=True
-        )
-        new_contours.save(
-            os.path.join(
-                output_dir, f"slice_{i:03d}.reg_{base_name}_inv.{base_name}_mask.vtp"
+            new_contours = con.transform_contours(
+                contours, forward_transform, with_deformation_magnitude=True
+            )
+            new_contours.save(
+                os.path.join(
+                    output_dir,
+                    f"slice_{i:03d}.reg_{base_name}_inv.{base_name}_mask.vtp",
+                )
             )
+
+    def convert_contours(base_name, output_dir, project_name, compute_normals=False):
+        files = [
+            f"{output_dir}/slice_{i:03d}.reg_{base_name}_inv.{base_name}_mask.vtp"
+            for i in range(0, N_FRAMES, FRAME_STEP)
+        ]
+        seg = SegmentChestTotalSegmentator()
+        all_mask_ids = seg.taxonomy.all_labels()
+
+        polydata = [pv.read(f) for f in files]
+
+        # For cardiac gated CT data with 21 time points (0-20), each frame = 1 second
+        # so we use times_per_second=1.0 instead of the default 24.0.
+        # Forwarding `seg` groups labels by anatomy type under
+        # /World/{project_name}/{type}/{label_name}.
+        converter = ConvertVTKToUSD(
+            project_name,
+            polydata,
+            all_mask_ids,
+            segmenter=seg,
+            compute_normals=compute_normals,
+            times_per_second=1.0,
+        )
+        stage = converter.convert(
+            os.path.join(output_dir, f"{project_name}.{base_name}.usd"),
         )
 
+        painter = USDAnatomyTools(stage)
+        painter.enhance_meshes(seg)
+        if os.path.exists(
+            os.path.join(output_dir, f"{project_name}.{base_name}_painted.usd")
+        ):
+            os.remove(
+                os.path.join(output_dir, f"{project_name}.{base_name}_painted.usd")
+            )
+        stage.Export(
+            os.path.join(output_dir, f"{project_name}.{base_name}_painted.usd")
+        )
 
-def convert_contours(base_name, output_dir, project_name, compute_normals=False):
-    files = [
-        f"{output_dir}/slice_{i:03d}.reg_{base_name}_inv.{base_name}_mask.vtp"
+    # %%
+    dynamic_transform_filenames = [
+        os.path.join(output_dir, f"slice_{i:03d}.reg_dynamic_anatomy.forward.hdf")
+        for i in range(0, N_FRAMES, FRAME_STEP)
+    ]
+    static_transform_filenames = [
+        os.path.join(output_dir, f"slice_{i:03d}.reg_static_anatomy.forward.hdf")
         for i in range(0, N_FRAMES, FRAME_STEP)
     ]
-    seg = SegmentChestTotalSegmentator()
-    all_mask_ids = seg.all_mask_ids
-
-    polydata = [pv.read(f) for f in files]
-
-    # For cardiac gated CT data with 21 time points (0-20), each frame = 1 second
-    # so we use times_per_second=1.0 instead of the default 24.0
-    converter = ConvertVTKToUSD(
-        project_name,
-        polydata,
-        all_mask_ids,
-        compute_normals=compute_normals,
-        times_per_second=1.0,
+    all_transform_filenames = [
+        os.path.join(output_dir, f"slice_{i:03d}.reg_all.forward.hdf")
+        for i in range(0, N_FRAMES, FRAME_STEP)
+    ]
+
+    dynamic_anatomy_contours = pv.read(
+        os.path.join(output_dir, f"{base_name}.dynamic_anatomy_mask.vtp")
     )
-    stage = converter.convert(
-        os.path.join(output_dir, f"{project_name}.{base_name}.usd"),
+    static_anatomy_contours = pv.read(
+        os.path.join(output_dir, f"{base_name}.static_anatomy_mask.vtp")
     )
+    all_contours = pv.read(os.path.join(output_dir, f"{base_name}.all_mask.vtp"))
 
-    painter = USDAnatomyTools(stage)
-    painter.enhance_meshes(seg)
-    if os.path.exists(
-        os.path.join(output_dir, f"{project_name}.{base_name}_painted.usd")
-    ):
-        os.remove(os.path.join(output_dir, f"{project_name}.{base_name}_painted.usd"))
-    stage.Export(os.path.join(output_dir, f"{project_name}.{base_name}_painted.usd"))
-
-
-# %%
-dynamic_transform_filenames = [
-    os.path.join(output_dir, f"slice_{i:03d}.reg_dynamic_anatomy.forward.hdf")
-    for i in range(0, N_FRAMES, FRAME_STEP)
-]
-static_transform_filenames = [
-    os.path.join(output_dir, f"slice_{i:03d}.reg_static_anatomy.forward.hdf")
-    for i in range(0, N_FRAMES, FRAME_STEP)
-]
-all_transform_filenames = [
-    os.path.join(output_dir, f"slice_{i:03d}.reg_all.forward.hdf")
-    for i in range(0, N_FRAMES, FRAME_STEP)
-]
-
-dynamic_anatomy_contours = pv.read(
-    os.path.join(output_dir, f"{base_name}.dynamic_anatomy_mask.vtp")
-)
-static_anatomy_contours = pv.read(
-    os.path.join(output_dir, f"{base_name}.static_anatomy_mask.vtp")
-)
-all_contours = pv.read(os.path.join(output_dir, f"{base_name}.all_mask.vtp"))
-
-frame_indices = list(range(0, N_FRAMES, FRAME_STEP))
+    frame_indices = list(range(0, N_FRAMES, FRAME_STEP))
 
-# %%
-transform_contours(
-    all_contours, all_transform_filenames, frame_indices, "all", output_dir
-)
-transform_contours(
-    dynamic_anatomy_contours,
-    dynamic_transform_filenames,
-    frame_indices,
-    "dynamic_anatomy",
-    output_dir,
-)
-transform_contours(
-    static_anatomy_contours,
-    static_transform_filenames,
-    frame_indices,
-    "static_anatomy",
-    output_dir,
-)
+    # %%
+    transform_contours(
+        all_contours, all_transform_filenames, frame_indices, "all", output_dir
+    )
+    transform_contours(
+        dynamic_anatomy_contours,
+        dynamic_transform_filenames,
+        frame_indices,
+        "dynamic_anatomy",
+        output_dir,
+    )
+    transform_contours(
+        static_anatomy_contours,
+        static_transform_filenames,
+        frame_indices,
+        "static_anatomy",
+        output_dir,
+    )
 
-# %%
-convert_contours("all", output_dir, project_name, compute_normals=True)
-convert_contours("dynamic_anatomy", output_dir, project_name, compute_normals=True)
-convert_contours("static_anatomy", output_dir, project_name, compute_normals=True)
+    # %%
+    convert_contours("all", output_dir, project_name, compute_normals=True)
+    convert_contours("dynamic_anatomy", output_dir, project_name, compute_normals=True)
+    convert_contours("static_anatomy", output_dir, project_name, compute_normals=True)
diff --git a/experiments/Heart-GatedCT_To_USD/test_compare_registration_speed.py b/experiments/Heart-GatedCT_To_USD/test_compare_registration_speed.py
index 6eb7c4f..95b9991 100644
--- a/experiments/Heart-GatedCT_To_USD/test_compare_registration_speed.py
+++ b/experiments/Heart-GatedCT_To_USD/test_compare_registration_speed.py
@@ -15,7 +15,7 @@
 import pandas as pd
 from itk import TubeTK as ttk
 
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 from physiomotion4d.register_images_ants import RegisterImagesANTs
 from physiomotion4d.register_images_greedy import RegisterImagesGreedy
 from physiomotion4d.register_images_icon import RegisterImagesICON
@@ -168,7 +168,7 @@
     ax.set_ylabel("Time (seconds)")
     ax.set_title("Registration time: two time points (Slicer-Heart-CT)")
     plt.tight_layout()
-    if not running_as_test():
+    if not TestTools.running_as_test():
         plt.show()
 else:
     print("No successful runs to plot.")
diff --git a/experiments/Heart-Simpleware_Segmentation/README.md b/experiments/Heart-Simpleware_Segmentation/README.md
index 2eb59a2..0dcf470 100644
--- a/experiments/Heart-Simpleware_Segmentation/README.md
+++ b/experiments/Heart-Simpleware_Segmentation/README.md
@@ -35,7 +35,7 @@ The `SegmentHeartSimpleware` class provides integration between PhysioMotion4D a
 
 ## Files
 
-- **`simpleware_heart_segmentation.ipynb`**: Main demonstration notebook
+- **`simpleware_heart_segmentation.py`**: Main demonstration script (percent-cell format)
 - **`README.md`**: This file
 - **`results/`**: Output directory (created automatically)
 
@@ -43,13 +43,16 @@ The `SegmentHeartSimpleware` class provides integration between PhysioMotion4D a
 
 ### Quick Start
 
-1. Open `simpleware_heart_segmentation.ipynb` in Jupyter
-2. Update the `input_image_path` in cell 3 to point to your cardiac CT image
-3. Run all cells sequentially
+1. Open `simpleware_heart_segmentation.py` in VS Code, Cursor, or any editor
+   with `# %%` cell support — or run it top-to-bottom with
+   `python simpleware_heart_segmentation.py`.
+2. Update the `input_image_path` near the top of the script to point to your
+   cardiac CT image.
+3. Run the cells sequentially or execute the file.
 
 ### Configuration
 
-Before running, configure these parameters in the notebook:
+Before running, configure these parameters in the script:
 
 ```python
 # Set input image path
@@ -61,7 +64,7 @@ custom_simpleware_path = "D:/CustomPath/Simpleware/ConsoleSimplewareMedical.exe"
 
 ### Expected Output
 
-The notebook generates:
+The script generates:
 
 1. **Segmentation files** (in `results/`):
    - `heart_labelmap_simpleware.nii.gz` - Complete labelmap with all structures
@@ -134,7 +137,7 @@ Times depend on image size, system performance, and Simpleware configuration.
 **Solution**:
 - Verify Simpleware installation
 - Ensure you're using `ConsoleSimplewareMedical.exe` (not `SimplewareMedical.exe`)
-- Update `custom_simpleware_path` in the notebook
+- Update `custom_simpleware_path` in the script
 - Check default path: `C:\Program Files\Synopsys\Simpleware Medical\X-2025.06\ConsoleSimplewareMedical.exe`; use `set_simpleware_executable_path()` if installed elsewhere
 
 ---
@@ -290,7 +293,7 @@ For issues with:
 ## Version History
 
 - **v0.2.0** (2026-02-06): Documentation and alignment with current implementation
-  - README reflects actual notebook name (`simpleware_heart_segmentation.ipynb`) and correct label IDs (heart 1–6, vessels 7–10)
+  - README reflects actual script name (`simpleware_heart_segmentation.py`) and correct label IDs (heart 1–6, vessels 7–10)
   - Workflow description updated for `--input-file` (NIfTI) and `--input-value` (output dir) invocation
   - Removed obsolete “placeholder output” limitation; integration works as expected
 - **v0.1.0** (2026-02-04): Initial implementation
diff --git a/experiments/Heart-Simpleware_Segmentation/simpleware_heart_segmentation.py b/experiments/Heart-Simpleware_Segmentation/simpleware_heart_segmentation.py
index 754f1d9..4f40d74 100644
--- a/experiments/Heart-Simpleware_Segmentation/simpleware_heart_segmentation.py
+++ b/experiments/Heart-Simpleware_Segmentation/simpleware_heart_segmentation.py
@@ -30,7 +30,7 @@
 import numpy as np
 import pyvista as pv
 
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 from physiomotion4d.segment_heart_simpleware import SegmentHeartSimpleware
 
 _HERE = os.path.dirname(os.path.abspath(__file__))
@@ -105,7 +105,7 @@
     axes[2].axis("off")
 
     plt.tight_layout()
-    if not running_as_test():
+    if not TestTools.running_as_test():
         plt.show()
 
     print(f"Image intensity range: [{image_array.min():.1f}, {image_array.max():.1f}]")
@@ -131,15 +131,17 @@
 # Display segmentation configuration
 print("\nSegmentation Configuration:")
 print(f"  Target spacing: {segmenter.target_spacing} mm")
-print(f"  Heart structures: {len(segmenter.heart_mask_ids)} labels")
-print(f"  Vessel structures: {len(segmenter.major_vessels_mask_ids)} labels")
+heart_labels = segmenter.taxonomy.labels_in_group("heart")
+vessel_labels = segmenter.taxonomy.labels_in_group("major_vessels")
+print(f"  Heart structures: {len(heart_labels)} labels")
+print(f"  Vessel structures: {len(vessel_labels)} labels")
 
 print("\nHeart Structure IDs:")
-for id, name in segmenter.heart_mask_ids.items():
+for id, name in heart_labels.items():
     print(f"  {id}: {name}")
 
 print("\nMajor Vessel IDs:")
-for id, name in segmenter.major_vessels_mask_ids.items():
+for id, name in vessel_labels.items():
     print(f"  {id}: {name}")
 
 # %% [markdown]
@@ -231,9 +233,9 @@
 
     # Combine all mask dictionaries for label lookup
     all_labels = {
-        **segmenter.heart_mask_ids,
-        **segmenter.major_vessels_mask_ids,
-        **segmenter.contrast_mask_ids,
+        **segmenter.taxonomy.labels_in_group("heart"),
+        **segmenter.taxonomy.labels_in_group("major_vessels"),
+        **segmenter.taxonomy.labels_in_group("contrast"),
     }
 
     for label, count in zip(unique_labels, label_counts):
@@ -336,7 +338,7 @@
 
     plt.tight_layout()
     plt.savefig(os.path.join(output_dir, "segmentation_visualization.png"), dpi=150)
-    if not running_as_test():
+    if not TestTools.running_as_test():
         plt.show()
 
     print(
@@ -391,7 +393,7 @@
 
     # Save screenshot
     screenshot_path = os.path.join(output_dir, "3d_visualization.png")
-    if not running_as_test():
+    if not TestTools.running_as_test():
         plotter.show(screenshot=screenshot_path)
 
     print(f"3D visualization saved to: {screenshot_path}")
diff --git a/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_model_icp_itk.py b/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_model_icp_itk.py
index b66e04c..92c2027 100644
--- a/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_model_icp_itk.py
+++ b/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_model_icp_itk.py
@@ -30,7 +30,7 @@
     RegisterModelsICPITK,
     TransformTools,
 )
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 
 # %% [markdown]
 # ## Define File Paths
@@ -204,5 +204,5 @@
 plotter.add_title("ICP Shape Fitting")
 plotter.add_axes()
 
-if not running_as_test():
+if not TestTools.running_as_test():
     plotter.show()
diff --git a/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_model_registration_pca.py b/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_model_registration_pca.py
index 48f62a2..18e4b1e 100644
--- a/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_model_registration_pca.py
+++ b/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_model_registration_pca.py
@@ -33,7 +33,7 @@
     RegisterModelsPCA,
     TransformTools,
 )
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 
 # %% [markdown]
 # ## Define File Paths
@@ -166,7 +166,7 @@
 icp_registrar = RegisterModelsICP(fixed_model=patient_surface)
 
 # Use fewer iterations when run as test (pytest) for faster execution
-max_iterations_icp = 100 if running_as_test() else 2000
+max_iterations_icp = 100 if TestTools.running_as_test() else 2000
 icp_result = icp_registrar.register(
     transform_type="Affine",
     moving_model=template_model_surface,
@@ -329,7 +329,7 @@
 plotter.add_axes()
 
 plotter.link_views()
-if not running_as_test():
+if not TestTools.running_as_test():
     plotter.show()
 
 # %% [markdown]
@@ -389,7 +389,7 @@
 )
 plotter.add_title("PCA Signed Displacement on Registered Model")
 plotter.add_axes()
-if not running_as_test():
+if not TestTools.running_as_test():
     plotter.show()
 
 # Save the mesh with displacement data
diff --git a/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_patient-CHOPValve.py b/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_patient-CHOPValve.py
index 1610c0f..0c00e15 100644
--- a/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_patient-CHOPValve.py
+++ b/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_patient-CHOPValve.py
@@ -14,7 +14,7 @@
 from physiomotion4d import (
     WorkflowFitStatisticalModelToPatient,
 )
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 
 # %% [markdown]
 # ## Define File Paths
@@ -114,7 +114,7 @@
 plotter.add_title("Final Registration")
 
 plotter.link_views()
-if not running_as_test():
+if not TestTools.running_as_test():
     plotter.show()
 
 # %% [markdown]
@@ -132,7 +132,7 @@
         scalar_bar_args={"title": "Deformation (mm)"},
     )
     plotter.add_title("Deformation Magnitude")
-    if not running_as_test():
+    if not TestTools.running_as_test():
         plotter.show()
 
     # Print statistics
diff --git a/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_patient.py b/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_patient.py
index 70e5fb7..7412633 100644
--- a/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_patient.py
+++ b/experiments/Heart-Statistical_Model_To_Patient/heart_model_to_patient.py
@@ -18,277 +18,299 @@
     SegmentChestTotalSegmentator,
     WorkflowFitStatisticalModelToPatient,
 )
-from physiomotion4d.notebook_utils import running_as_test
-
-# %% [markdown]
-# ## Define File Paths
-
-# %%
-# Patient CT image (defines coordinate frame)
-patient_data_dir = Path.cwd().parent / ".." / "data" / "Slicer-Heart-CT"
-patient_ct_path = patient_data_dir / "patient_img.mha"
-patient_ct_heart_mask_path = patient_data_dir / "patient_heart_wall_mask.nii.gz"
-
-# Atlas template model (moving)
-atlas_data_dir = Path.cwd().parent / ".." / "data" / "KCL-Heart-Model"
-atlas_vtu_path = atlas_data_dir / "average_mesh.vtk"
-atlas_labelmap_path = atlas_data_dir / "labelmap" / "average_labelmap_with_bkg.mha"
-
-pca_data_dir = Path.cwd().parent / "Heart-Create_Statistical_Model" / "kcl-heart-model"
-pca_json_path = pca_data_dir / "pca_model.json"
-pca_n_modes = 10
-
-# Output directory
-output_dir = Path.cwd() / "results"
-
-os.makedirs(output_dir, exist_ok=True)
-
-# %%
-patient_image = itk.imread(str(patient_ct_path))
-itk.imwrite(patient_image, str(output_dir / "patient_image.mha"), compression=True)
-
-# %%
-if False:
-    segmentator = SegmentChestTotalSegmentator()
-    segmentator.contrast_threshold = 500
-    patient_segmentation_data = segmentator.segment(
-        patient_image, contrast_enhanced_study=False
+from physiomotion4d.test_tools import TestTools
+
+# nnUNetv2 (used by TotalSegmentator) spawns a multiprocessing.Pool. On Windows
+# the spawn start method re-imports this script in each child; without the
+# __name__ == "__main__" guard around the top-level work, that re-import fires
+# segment() again and Python's spawn-cascade detector raises RuntimeError. The
+# segmenter call below is currently gated by `if False:` but the guard is kept
+# in case that branch is re-enabled, and to be consistent with sibling scripts.
+if __name__ == "__main__":
+    # %% [markdown]
+    # ## Define File Paths
+
+    # %%
+    # Patient CT image (defines coordinate frame)
+    patient_data_dir = Path.cwd().parent / ".." / "data" / "Slicer-Heart-CT"
+    patient_ct_path = patient_data_dir / "patient_img.mha"
+    patient_ct_heart_mask_path = patient_data_dir / "patient_heart_wall_mask.nii.gz"
+
+    # Atlas template model (moving)
+    atlas_data_dir = Path.cwd().parent / ".." / "data" / "KCL-Heart-Model"
+    atlas_vtu_path = atlas_data_dir / "average_mesh.vtk"
+    atlas_labelmap_path = atlas_data_dir / "labelmap" / "average_labelmap_with_bkg.mha"
+
+    pca_data_dir = (
+        Path.cwd().parent / "Heart-Create_Statistical_Model" / "kcl-heart-model"
     )
-    labelmap = patient_segmentation_data["labelmap"]
-    lung_mask = patient_segmentation_data["lung"]
-    heart_mask = patient_segmentation_data["heart"]
-    major_vessels_mask = patient_segmentation_data["major_vessels"]
-    bone_mask = patient_segmentation_data["bone"]
-    soft_tissue_mask = patient_segmentation_data["soft_tissue"]
-    other_mask = patient_segmentation_data["other"]
-    contrast_mask = patient_segmentation_data["contrast"]
-
-    itk.imwrite(labelmap, str(output_dir / "patient_labelmap.mha"), compression=True)
-
-    heart_arr = itk.GetArrayFromImage(heart_mask)
-    # contrast_arr = itk.GetArrayFromImage(contrast_mask)
-    mask_arr = (heart_arr > 0).astype(
-        np.uint8
-    )  # ((heart_arr + contrast_arr) > 0).astype(np.uint8)
-    patient_mask = itk.GetImageFromArray(mask_arr)
-    patient_mask.CopyInformation(patient_image)
+    pca_json_path = pca_data_dir / "pca_model.json"
+    pca_n_modes = 10
 
-    itk.imwrite(
-        patient_mask, str(output_dir / "patient_heart_mask_draft.mha"), compression=True
-    )
+    # Output directory
+    output_dir = Path.cwd() / "results"
 
-    # hand edit fixed_mask to make patient_heart_wall_mask.nii.gz that is saved in patient_data_dir
-else:
-    patient_mask = itk.imread(str(patient_ct_heart_mask_path))
+    os.makedirs(output_dir, exist_ok=True)
 
-# %%
-flip0 = np.array(patient_mask.GetDirection())[0, 0] < 0
-flip1 = np.array(patient_mask.GetDirection())[1, 1] < 0
-flip2 = np.array(patient_mask.GetDirection())[2, 2] < 0
-if flip0 or flip1 or flip2:
-    print("Flipping patient image...")
-    print(flip0, flip1, flip2)
-    flip_filter = itk.FlipImageFilter.New(Input=patient_image)
-    flip_filter.SetFlipAxes([int(flip0), int(flip1), int(flip2)])
-    flip_filter.SetFlipAboutOrigin(True)
-    flip_filter.Update()
-    patient_image = flip_filter.GetOutput()
-    id_mat = itk.Matrix[itk.D, 3, 3]()
-    id_mat.SetIdentity()
-    patient_image.SetDirection(id_mat)
+    # %%
+    patient_image = itk.imread(str(patient_ct_path))
     itk.imwrite(patient_image, str(output_dir / "patient_image.mha"), compression=True)
-    print("Flipping patient mask image...")
-    flip_filter = itk.FlipImageFilter.New(Input=patient_mask)
-    flip_filter.SetFlipAxes([int(flip0), int(flip1), int(flip2)])
-    flip_filter.SetFlipAboutOrigin(True)
-    flip_filter.Update()
-    patient_mask = flip_filter.GetOutput()
-    patient_mask.SetDirection(id_mat)
-    itk.imwrite(patient_mask, str(output_dir / "patient_mask.mha"), compression=True)
-
-# %%
-patient_model = ContourTools().extract_contours(patient_mask)
-patient_model.save(str(output_dir / "patient_mesh.vtp"))
-patient_model = pv.read(str(output_dir / "patient_mesh.vtp"))
-
-template_model = pv.read(str(atlas_vtu_path))
-template_model_surface = template_model.extract_surface(algorithm="dataset_surface")
-template_model_surface.save(str(output_dir / "model_surface.vtp"))
-template_model_surface = pv.read(str(output_dir / "model_surface.vtp"))
-template_labelmap = itk.imread(str(atlas_labelmap_path))
-
-# %%
-with open(pca_json_path, encoding="utf-8") as f:
-    pca_model = json.load(f)
-registrar = WorkflowFitStatisticalModelToPatient(
-    template_model=template_model,
-    patient_models=[patient_model],
-    patient_image=patient_image,
-)
-registrar.set_use_pca_registration(
-    True, pca_model=pca_model, pca_number_of_modes=pca_n_modes
-)
-registrar.set_use_mask_to_image_registration(
-    True,
-    template_labelmap=template_labelmap,
-    template_labelmap_organ_mesh_ids=[1],
-    template_labelmap_organ_extra_ids=[2, 3, 4, 5],
-    template_labelmap_background_ids=[6],
-)
-
-registrar.set_mask_dilation_mm(0)
-registrar.set_roi_dilation_mm(25)
-
-patient_image = registrar.patient_image
-itk.imwrite(
-    patient_image, str(output_dir / "patient_image_preprocessed.mha"), compression=True
-)
-
-# %%
-# Rough alignment using ICP
-icp_results = registrar.register_model_to_model_icp()
-icp_inverse_point_transform = icp_results["inverse_point_transform"]
-icp_forward_point_transform = icp_results["forward_point_transform"]
-icp_model_surface = icp_results["registered_template_model_surface"]
-icp_labelmap = icp_results["registered_template_labelmap"]
-
-icp_model_surface.save(str(output_dir / "icp_model_surface.vtp"))
-itk.imwrite(icp_labelmap, str(output_dir / "icp_labelmap.mha"), compression=True)
 
-# %%
-pca_results = registrar.register_model_to_model_pca()
-pca_coefficients = pca_results["pca_coefficients"]
-pca_model_surface = pca_results["registered_template_model_surface"]
-pca_labelmap = pca_results["registered_template_labelmap"]
-
-pca_model_surface.save(str(output_dir / "pca_model_surface.vtp"))
-itk.imwrite(pca_labelmap, str(output_dir / "pca_labelmap.mha"), compression=True)
-
-# %% [markdown]
-# ## Mask Alignment
-
-# %%
-# Perform deformable registration
-print("Starting deformable mask-to-mask registration...")
-
-m2m_results = registrar.register_mask_to_mask(use_icon_refinement=False)
-m2m_inverse_transform = m2m_results["inverse_transform"]
-m2m_forward_transform = m2m_results["forward_transform"]
-m2m_model_surface = m2m_results["registered_template_model_surface"]
-m2m_labelmap = m2m_results["registered_template_labelmap"]
-
-print("Registration complete!")
-
-m2m_model_surface.save(str(output_dir / "m2m_model_surface.vtp"))
-itk.imwrite(m2m_labelmap, str(output_dir / "m2m_labelmap.mha"), compression=True)
-
-# %%
-print("Starting deformable registration...")
-print("This may take several minutes depending on GPU availability.")
-
-m2i_results = registrar.register_labelmap_to_image()
-m2i_inverse_transform = m2i_results["inverse_transform"]
-m2i_forward_transform = m2i_results["forward_transform"]
-m2i_surface = m2i_results["registered_template_model_surface"]
-m2i_labelmap = m2i_results["registered_template_labelmap"]
-print("\nRegistration complete!")
-
-# Save registration results to output folder
-m2i_surface.save(str(output_dir / "m2i_model_surface.vtp"))
-itk.imwrite(m2i_labelmap, str(output_dir / "m2i_labelmap.mha"), compression=True)
-
-# %%
-tmp_p = itk.Point[itk.D, 3]()
-point = registrar.template_model.points[0]
-tmp_p[0] = float(point[0])
-tmp_p[1] = float(point[1])
-tmp_p[2] = float(point[2])
-
-start_time = time.time()
-# Timing only; the workflow stores the ICP-aligned model before PCA registration.
-icp_p = registrar.icp_registrar.forward_point_transform.TransformPoint(tmp_p)
-print(f"--- ICP forward transform time: {time.time() - start_time} seconds", flush=True)
-
-start_time = time.time()
-# Timing only; PCA registration operates in the ICP-aligned coordinate frame here.
-_ = registrar.pca_registrar.transform_point(icp_p)
-print(f"--- PCA setup time: {time.time() - start_time} seconds", flush=True)
-start_time = time.time()
-# Apply the PCA transform in the workflow's ICP-aligned coordinate frame.
-tmp_p = registrar.pca_registrar.transform_point(icp_p)
-print(f"PCA transform time: {time.time() - start_time} seconds", flush=True)
-
-start_time = time.time()
-tmp_p = registrar.m2m_inverse_transform.TransformPoint(tmp_p)
-print(f"M2M inverse transform time: {time.time() - start_time} seconds", flush=True)
-
-start_time = time.time()
-tmp_p = registrar.m2i_inverse_transform.TransformPoint(tmp_p)
-print(f"M2I inverse transform time: {time.time() - start_time} seconds", flush=True)
-
-# %%
-# Verify registration using the transform member function
-surface_transformed = registrar.m2i_template_model_surface
-surface_transformed.save(str(output_dir / "registered_template_surface.vtp"))
-
-model_transformed = registrar.transform_model()
-model_transformed.save(str(output_dir / "registered_template.vtu"))
-
-# %% [markdown]
-# ## Visualize Final Results
+    # %%
+    if False:
+        segmentator = SegmentChestTotalSegmentator()
+        segmentator.contrast_threshold = 500
+        patient_segmentation_data = segmentator.segment(
+            patient_image, contrast_enhanced_study=False
+        )
+        labelmap = patient_segmentation_data["labelmap"]
+        lung_mask = patient_segmentation_data["lung"]
+        heart_mask = patient_segmentation_data["heart"]
+        major_vessels_mask = patient_segmentation_data["major_vessels"]
+        bone_mask = patient_segmentation_data["bone"]
+        soft_tissue_mask = patient_segmentation_data["soft_tissue"]
+        other_mask = patient_segmentation_data["other"]
+        contrast_mask = patient_segmentation_data["contrast"]
+
+        itk.imwrite(
+            labelmap, str(output_dir / "patient_labelmap.mha"), compression=True
+        )
+
+        heart_arr = itk.GetArrayFromImage(heart_mask)
+        # contrast_arr = itk.GetArrayFromImage(contrast_mask)
+        mask_arr = (heart_arr > 0).astype(
+            np.uint8
+        )  # ((heart_arr + contrast_arr) > 0).astype(np.uint8)
+        patient_mask = itk.GetImageFromArray(mask_arr)
+        patient_mask.CopyInformation(patient_image)
+
+        itk.imwrite(
+            patient_mask,
+            str(output_dir / "patient_heart_mask_draft.mha"),
+            compression=True,
+        )
+
+        # hand edit fixed_mask to make patient_heart_wall_mask.nii.gz that is saved in patient_data_dir
+    else:
+        patient_mask = itk.imread(str(patient_ct_heart_mask_path))
+
+    # %%
+    flip0 = np.array(patient_mask.GetDirection())[0, 0] < 0
+    flip1 = np.array(patient_mask.GetDirection())[1, 1] < 0
+    flip2 = np.array(patient_mask.GetDirection())[2, 2] < 0
+    if flip0 or flip1 or flip2:
+        print("Flipping patient image...")
+        print(flip0, flip1, flip2)
+        flip_filter = itk.FlipImageFilter.New(Input=patient_image)
+        flip_filter.SetFlipAxes([int(flip0), int(flip1), int(flip2)])
+        flip_filter.SetFlipAboutOrigin(True)
+        flip_filter.Update()
+        patient_image = flip_filter.GetOutput()
+        id_mat = itk.Matrix[itk.D, 3, 3]()
+        id_mat.SetIdentity()
+        patient_image.SetDirection(id_mat)
+        itk.imwrite(
+            patient_image, str(output_dir / "patient_image.mha"), compression=True
+        )
+        print("Flipping patient mask image...")
+        flip_filter = itk.FlipImageFilter.New(Input=patient_mask)
+        flip_filter.SetFlipAxes([int(flip0), int(flip1), int(flip2)])
+        flip_filter.SetFlipAboutOrigin(True)
+        flip_filter.Update()
+        patient_mask = flip_filter.GetOutput()
+        patient_mask.SetDirection(id_mat)
+        itk.imwrite(
+            patient_mask, str(output_dir / "patient_mask.mha"), compression=True
+        )
+
+    # %%
+    patient_model = ContourTools().extract_contours(patient_mask)
+    patient_model.save(str(output_dir / "patient_mesh.vtp"))
+    patient_model = pv.read(str(output_dir / "patient_mesh.vtp"))
+
+    template_model = pv.read(str(atlas_vtu_path))
+    template_model_surface = template_model.extract_surface(algorithm="dataset_surface")
+    template_model_surface.save(str(output_dir / "model_surface.vtp"))
+    template_model_surface = pv.read(str(output_dir / "model_surface.vtp"))
+    template_labelmap = itk.imread(str(atlas_labelmap_path))
+
+    # %%
+    with open(pca_json_path, encoding="utf-8") as f:
+        pca_model = json.load(f)
+    registrar = WorkflowFitStatisticalModelToPatient(
+        template_model=template_model,
+        patient_models=[patient_model],
+        patient_image=patient_image,
+    )
+    registrar.set_use_pca_registration(
+        True, pca_model=pca_model, pca_number_of_modes=pca_n_modes
+    )
+    registrar.set_use_mask_to_image_registration(
+        True,
+        template_labelmap=template_labelmap,
+        template_labelmap_organ_mesh_ids=[1],
+        template_labelmap_organ_extra_ids=[2, 3, 4, 5],
+        template_labelmap_background_ids=[6],
+    )
 
-# %%
-# Load meshes from registrar member variables
-patient_surface = registrar.patient_model_surface
-registered_surface = registrar.registered_template_model_surface
-icp_surface = registrar.icp_template_model_surface
-pca_surface = registrar.pca_template_model_surface
-m2m_surface = registrar.m2m_template_model_surface
-m2i_surface = registrar.m2i_template_model_surface
-
-# Create side-by-side comparison
-plotter = pv.Plotter(shape=(1, 2))
-
-# After rough alignment
-plotter.subplot(0, 0)
-plotter.add_mesh(patient_surface, color="red", opacity=0.5, label="Patient")
-plotter.add_mesh(pca_surface, color="green", opacity=1.0, label="After ICP")
-plotter.add_title("PCA Alignment")
-
-# After deformable registration
-plotter.subplot(0, 1)
-plotter.add_mesh(patient_surface, color="red", opacity=0.5, label="Patient")
-plotter.add_mesh(m2i_surface, color="blue", opacity=1.0, label="Registered")
-plotter.add_title("Final Registration")
-
-plotter.link_views()
-if not running_as_test():
-    plotter.show()
+    registrar.set_mask_dilation_mm(0)
+    registrar.set_roi_dilation_mm(25)
 
-# %% [markdown]
-# ## Visualize Deformation Magnitude
+    patient_image = registrar.patient_image
+    itk.imwrite(
+        patient_image,
+        str(output_dir / "patient_image_preprocessed.mha"),
+        compression=True,
+    )
 
-# %%
-# The transformed mesh has deformation magnitude stored as point data
-if "DeformationMagnitude" in registered_surface.point_data:
-    plotter = pv.Plotter()
-    plotter.add_mesh(
-        registered_surface,
-        scalars="DeformationMagnitude",
-        cmap="jet",
-        show_scalar_bar=True,
-        scalar_bar_args={"title": "Deformation (mm)"},
+    # %%
+    # Rough alignment using ICP
+    icp_results = registrar.register_model_to_model_icp()
+    icp_inverse_point_transform = icp_results["inverse_point_transform"]
+    icp_forward_point_transform = icp_results["forward_point_transform"]
+    icp_model_surface = icp_results["registered_template_model_surface"]
+    icp_labelmap = icp_results["registered_template_labelmap"]
+
+    icp_model_surface.save(str(output_dir / "icp_model_surface.vtp"))
+    itk.imwrite(icp_labelmap, str(output_dir / "icp_labelmap.mha"), compression=True)
+
+    # %%
+    pca_results = registrar.register_model_to_model_pca()
+    pca_coefficients = pca_results["pca_coefficients"]
+    pca_model_surface = pca_results["registered_template_model_surface"]
+    pca_labelmap = pca_results["registered_template_labelmap"]
+
+    pca_model_surface.save(str(output_dir / "pca_model_surface.vtp"))
+    itk.imwrite(pca_labelmap, str(output_dir / "pca_labelmap.mha"), compression=True)
+
+    # %% [markdown]
+    # ## Mask Alignment
+
+    # %%
+    # Perform deformable registration
+    print("Starting deformable mask-to-mask registration...")
+
+    m2m_results = registrar.register_mask_to_mask(use_icon_refinement=False)
+    m2m_inverse_transform = m2m_results["inverse_transform"]
+    m2m_forward_transform = m2m_results["forward_transform"]
+    m2m_model_surface = m2m_results["registered_template_model_surface"]
+    m2m_labelmap = m2m_results["registered_template_labelmap"]
+
+    print("Registration complete!")
+
+    m2m_model_surface.save(str(output_dir / "m2m_model_surface.vtp"))
+    itk.imwrite(m2m_labelmap, str(output_dir / "m2m_labelmap.mha"), compression=True)
+
+    # %%
+    print("Starting deformable registration...")
+    print("This may take several minutes depending on GPU availability.")
+
+    m2i_results = registrar.register_labelmap_to_image()
+    m2i_inverse_transform = m2i_results["inverse_transform"]
+    m2i_forward_transform = m2i_results["forward_transform"]
+    m2i_surface = m2i_results["registered_template_model_surface"]
+    m2i_labelmap = m2i_results["registered_template_labelmap"]
+    print("\nRegistration complete!")
+
+    # Save registration results to output folder
+    m2i_surface.save(str(output_dir / "m2i_model_surface.vtp"))
+    itk.imwrite(m2i_labelmap, str(output_dir / "m2i_labelmap.mha"), compression=True)
+
+    # %%
+    tmp_p = itk.Point[itk.D, 3]()
+    point = registrar.template_model.points[0]
+    tmp_p[0] = float(point[0])
+    tmp_p[1] = float(point[1])
+    tmp_p[2] = float(point[2])
+
+    start_time = time.time()
+    # Timing only; the workflow stores the ICP-aligned model before PCA registration.
+    icp_p = registrar.icp_registrar.forward_point_transform.TransformPoint(tmp_p)
+    print(
+        f"--- ICP forward transform time: {time.time() - start_time} seconds",
+        flush=True,
     )
-    plotter.add_title("Deformation Magnitude")
-    if not running_as_test():
+
+    start_time = time.time()
+    # Timing only; PCA registration operates in the ICP-aligned coordinate frame here.
+    _ = registrar.pca_registrar.transform_point(icp_p)
+    print(f"--- PCA setup time: {time.time() - start_time} seconds", flush=True)
+    start_time = time.time()
+    # Apply the PCA transform in the workflow's ICP-aligned coordinate frame.
+    tmp_p = registrar.pca_registrar.transform_point(icp_p)
+    print(f"PCA transform time: {time.time() - start_time} seconds", flush=True)
+
+    start_time = time.time()
+    tmp_p = registrar.m2m_inverse_transform.TransformPoint(tmp_p)
+    print(f"M2M inverse transform time: {time.time() - start_time} seconds", flush=True)
+
+    start_time = time.time()
+    tmp_p = registrar.m2i_inverse_transform.TransformPoint(tmp_p)
+    print(f"M2I inverse transform time: {time.time() - start_time} seconds", flush=True)
+
+    # %%
+    # Verify registration using the transform member function
+    surface_transformed = registrar.m2i_template_model_surface
+    surface_transformed.save(str(output_dir / "registered_template_surface.vtp"))
+
+    model_transformed = registrar.transform_model()
+    model_transformed.save(str(output_dir / "registered_template.vtu"))
+
+    # %% [markdown]
+    # ## Visualize Final Results
+
+    # %%
+    # Load meshes from registrar member variables
+    patient_surface = registrar.patient_model_surface
+    registered_surface = registrar.registered_template_model_surface
+    icp_surface = registrar.icp_template_model_surface
+    pca_surface = registrar.pca_template_model_surface
+    m2m_surface = registrar.m2m_template_model_surface
+    m2i_surface = registrar.m2i_template_model_surface
+
+    # Create side-by-side comparison
+    plotter = pv.Plotter(shape=(1, 2))
+
+    # After rough alignment
+    plotter.subplot(0, 0)
+    plotter.add_mesh(patient_surface, color="red", opacity=0.5, label="Patient")
+    plotter.add_mesh(pca_surface, color="green", opacity=1.0, label="After ICP")
+    plotter.add_title("PCA Alignment")
+
+    # After deformable registration
+    plotter.subplot(0, 1)
+    plotter.add_mesh(patient_surface, color="red", opacity=0.5, label="Patient")
+    plotter.add_mesh(m2i_surface, color="blue", opacity=1.0, label="Registered")
+    plotter.add_title("Final Registration")
+
+    plotter.link_views()
+    if not TestTools.running_as_test():
         plotter.show()
 
-    # Print statistics
-    deformation = registered_surface["DeformationMagnitude"]
-    print("Deformation statistics:")
-    print(f"  Min: {deformation.min():.2f} mm")
-    print(f"  Max: {deformation.max():.2f} mm")
-    print(f"  Mean: {deformation.mean():.2f} mm")
-    print(f"  Std: {deformation.std():.2f} mm")
-else:
-    print("DeformationMagnitude not found in mesh point data")
+    # %% [markdown]
+    # ## Visualize Deformation Magnitude
+
+    # %%
+    # The transformed mesh has deformation magnitude stored as point data
+    if "DeformationMagnitude" in registered_surface.point_data:
+        plotter = pv.Plotter()
+        plotter.add_mesh(
+            registered_surface,
+            scalars="DeformationMagnitude",
+            cmap="jet",
+            show_scalar_bar=True,
+            scalar_bar_args={"title": "Deformation (mm)"},
+        )
+        plotter.add_title("Deformation Magnitude")
+        if not TestTools.running_as_test():
+            plotter.show()
+
+        # Print statistics
+        deformation = registered_surface["DeformationMagnitude"]
+        print("Deformation statistics:")
+        print(f"  Min: {deformation.min():.2f} mm")
+        print(f"  Max: {deformation.max():.2f} mm")
+        print(f"  Mean: {deformation.mean():.2f} mm")
+        print(f"  Std: {deformation.std():.2f} mm")
+    else:
+        print("DeformationMagnitude not found in mesh point data")
diff --git a/experiments/Heart-VTKSeries_To_USD/0-download_and_convert_4d_to_3d.py b/experiments/Heart-VTKSeries_To_USD/0-download_and_convert_4d_to_3d.py
index fcebe0f..dfc0eb2 100644
--- a/experiments/Heart-VTKSeries_To_USD/0-download_and_convert_4d_to_3d.py
+++ b/experiments/Heart-VTKSeries_To_USD/0-download_and_convert_4d_to_3d.py
@@ -1,9 +1,9 @@
 #!/usr/bin/env python
 # %%
 import os
-import urllib.request
 
 from physiomotion4d.convert_nrrd_4d_to_3d import ConvertNRRD4DTo3D
+from physiomotion4d.data_download_tools import DataDownloadTools
 
 _HERE = os.path.dirname(os.path.abspath(__file__))
 
@@ -17,15 +17,10 @@
 if not os.path.exists(output_dir):
     os.makedirs(output_dir)
 
-# %%
-input_image_url = "https://github.com/SlicerHeart/SlicerHeart/releases/download/TestingData/TruncalValve_4DCT.seq.nrrd"
-input_image_filename = os.path.join(data_dir, "TruncalValve_4DCT.seq.nrrd")
-
-if not os.path.exists(input_image_filename):
-    urllib.request.urlretrieve(input_image_url, input_image_filename)
+input_image_filename = DataDownloadTools.DownloadSlicerHeartCTData(data_dir)
 
 # %%
 if not os.path.exists(f"{data_dir}/slice_000.mha"):
     conv = ConvertNRRD4DTo3D()
-    conv.load_nrrd_4d(f"{data_dir}/TruncalValve_4DCT.seq.nrrd")
-    conv.save_3d_images(f"{data_dir}/slice")
+    conv.load_nrrd_4d(str(input_image_filename))
+    conv.save_3d_images(data_dir, "slice")
diff --git a/experiments/Heart-VTKSeries_To_USD/1-heart_vtkseries_to_usd.py b/experiments/Heart-VTKSeries_To_USD/1-heart_vtkseries_to_usd.py
index 36c8cee..4fc042f 100644
--- a/experiments/Heart-VTKSeries_To_USD/1-heart_vtkseries_to_usd.py
+++ b/experiments/Heart-VTKSeries_To_USD/1-heart_vtkseries_to_usd.py
@@ -7,62 +7,67 @@
 
 from physiomotion4d import ConvertVTKToUSD
 
-_HERE = os.path.dirname(os.path.abspath(__file__))
-_DATA_DIR = os.path.join(_HERE, "..", "..", "data", "Slicer-Heart-CT")
+# nnUNetv2 (used by TotalSegmentator) spawns a multiprocessing.Pool. On Windows
+# the spawn start method re-imports this script in each child; without the
+# __name__ == "__main__" guard around the top-level work, that re-import fires
+# segment() again and Python's spawn-cascade detector raises RuntimeError.
+if __name__ == "__main__":
+    _HERE = os.path.dirname(os.path.abspath(__file__))
+    _DATA_DIR = os.path.join(_HERE, "..", "..", "data", "Slicer-Heart-CT")
 
-# %%
-if not os.path.exists(os.path.join(_DATA_DIR, "slice_000.vtp")):
-    # Segment chest from CT images to generate vtk files
-    import itk
+    # %%
+    if not os.path.exists(os.path.join(_DATA_DIR, "slice_000.vtp")):
+        # Segment chest from CT images to generate vtk files
+        import itk
 
-    from physiomotion4d.contour_tools import ContourTools
-    from physiomotion4d.segment_chest_total_segmentator import (
-        SegmentChestTotalSegmentator,
-    )
+        from physiomotion4d.contour_tools import ContourTools
+        from physiomotion4d.segment_chest_total_segmentator import (
+            SegmentChestTotalSegmentator,
+        )
 
-    input_images = sorted(glob.glob(os.path.join(_DATA_DIR, "slice_*.mha")))
-    seg = SegmentChestTotalSegmentator()
-    seg.contrast_threshold = 500
-    con = ContourTools()
-    for i, img_path in enumerate(input_images):
-        print(f"Segmenting {img_path}...")
-        img = itk.imread(img_path)
-        result = seg.segment(img, contrast_enhanced_study=True)
-        labelmap_mask = result["labelmap"]
-        img_con = con.extract_contours(labelmap_mask)
-        img_con.save(os.path.join(_DATA_DIR, f"slice_{i:03d}.vtp"))
+        input_images = sorted(glob.glob(os.path.join(_DATA_DIR, "slice_*.mha")))
+        seg = SegmentChestTotalSegmentator()
+        seg.contrast_threshold = 500
+        con = ContourTools()
+        for i, img_path in enumerate(input_images):
+            print(f"Segmenting {img_path}...")
+            img = itk.imread(img_path)
+            result = seg.segment(img, contrast_enhanced_study=True)
+            labelmap_mask = result["labelmap"]
+            img_con = con.extract_contours(labelmap_mask)
+            img_con.save(os.path.join(_DATA_DIR, f"slice_{i:03d}.vtp"))
 
-# %%
-project_name = "Heart_VTKSeries_To_USD_all"
+    # %%
+    project_name = "Heart_VTKSeries_To_USD_all"
 
-input_files = sorted(glob.glob(os.path.join(_DATA_DIR, "slice_*.vtp")))
+    input_files = sorted(glob.glob(os.path.join(_DATA_DIR, "slice_*.vtp")))
 
-output_dir = os.path.join(_HERE, "results")
-os.makedirs(output_dir, exist_ok=True)
+    output_dir = os.path.join(_HERE, "results")
+    os.makedirs(output_dir, exist_ok=True)
 
-input_polydata = []
-for file in input_files:
-    print(f"Processing file: {file}")
-    pd = pv.read(file)
-    print(f"  Number of points: {pd.n_points}")
-    # Print available data arrays
-    if len(input_polydata) == 0:  # Only print for first file
-        print(f"  Point data arrays: {list(pd.point_data.keys())}")
-    input_polydata.append(pd)
+    input_polydata = []
+    for file in input_files:
+        print(f"Processing file: {file}")
+        pd = pv.read(file)
+        print(f"  Number of points: {pd.n_points}")
+        # Print available data arrays
+        if len(input_polydata) == 0:  # Only print for first file
+            print(f"  Point data arrays: {list(pd.point_data.keys())}")
+        input_polydata.append(pd)
 
-# Convert with transmembrane potential coloring
-converter = ConvertVTKToUSD(
-    project_name,
-    input_polydata,
-)
-# converter.set_colormap(
-# color_by_array='transmembrane_potential',
-# colormap='viridis',
-# intensity_range=(-80.0, 40.0),
-# )
-stage = converter.convert(
-    os.path.join(output_dir, f"{project_name}.usd"), convert_to_surface=True
-)
+    # Convert with transmembrane potential coloring
+    converter = ConvertVTKToUSD(
+        project_name,
+        input_polydata,
+    )
+    # converter.set_colormap(
+    # color_by_array='transmembrane_potential',
+    # colormap='viridis',
+    # intensity_range=(-80.0, 40.0),
+    # )
+    stage = converter.convert(
+        os.path.join(output_dir, f"{project_name}.usd"), convert_to_surface=True
+    )
 
-print("\nUSD files created!")
-print(f"  - {os.path.join(output_dir, f'{project_name}.usd')}")
+    print("\nUSD files created!")
+    print(f"  - {os.path.join(output_dir, f'{project_name}.usd')}")
diff --git a/experiments/Lung-GatedCT_To_USD/0-register_dirlab_4dct.py b/experiments/Lung-GatedCT_To_USD/0-register_dirlab_4dct.py
index 2a08ad7..ef3cbb5 100644
--- a/experiments/Lung-GatedCT_To_USD/0-register_dirlab_4dct.py
+++ b/experiments/Lung-GatedCT_To_USD/0-register_dirlab_4dct.py
@@ -12,275 +12,281 @@
 from physiomotion4d.segment_chest_total_segmentator import SegmentChestTotalSegmentator
 from physiomotion4d.transform_tools import TransformTools
 
-_HERE = os.path.dirname(os.path.abspath(__file__))
+# nnUNetv2 (used by TotalSegmentator) spawns a multiprocessing.Pool. On Windows
+# the spawn start method re-imports this script in each child; without the
+# __name__ == "__main__" guard around the top-level work, that re-import fires
+# segment() again and Python's spawn-cascade detector raises RuntimeError.
+if __name__ == "__main__":
+    _HERE = os.path.dirname(os.path.abspath(__file__))
+
+    fixed_image_num = 3
+    heart_mask_dilation = 5
+
+    case_names = DataDirLab4DCT().case_names
+    case_names = [case_names[4]]
+    images = range(10)
+    # images = [1]
+
+    input_dir = os.path.join(_HERE, "..", "..", "data", "DirLab-4DCT")
+    output_dir = os.path.join(_HERE, "results")
+
+    # %%
+    def dilate_mask(mask: Optional[itk.image], dilation: int) -> Optional[itk.image]:
+        if mask is not None:
+            im_math = tube.ImageMath.New(mask)
+            im_math.Dilate(dilation, 1, 0)
+            dilated_mask = im_math.GetOutputShort()
+            return dilated_mask
+        return None
+
+    def register_image(
+        fixed_image: itk.image,
+        fixed_mask: itk.image,
+        moving_image: itk.image,
+        moving_mask: itk.image,
+        case_name: str,
+        image_num: int,
+        mask_name: str,
+        output_dir: str,
+    ) -> None:
+        """
+        Register a moving image to a fixed image using a mask.
+        """
+
+        reg_images = RegisterImagesICON()
+        reg_images.set_modality("ct")
+        reg_images.set_number_of_iterations(20)
+
+        if moving_mask is not None:
+            itk.imwrite(
+                moving_mask,
+                f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_mask_org.mha",
+                compression=True,
+            )
 
-fixed_image_num = 3
-heart_mask_dilation = 5
+        print("Registering image...")
+        reg_images.set_fixed_image(fixed_image)
+        moving_mask_d = None
+        if fixed_mask is not None:
+            fixed_mask_d = dilate_mask(fixed_mask, heart_mask_dilation)
+            moving_mask_d = dilate_mask(moving_mask, heart_mask_dilation)
+            reg_images.set_fixed_mask(fixed_mask_d)
+        results = reg_images.register(moving_image, moving_mask_d)
+        inverse_transform = results["inverse_transform"]
+        forward_transform = results["forward_transform"]
+        print("Registering image...Done!")
+        moving_image_reg = TransformTools().transform_image(
+            moving_image, forward_transform, fixed_image, "sinc"
+        )  # Final resampling with sinc
+        itk.imwrite(
+            moving_image_reg,
+            f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_reg.mha",
+            compression=True,
+        )
 
-case_names = DataDirLab4DCT().case_names
-case_names = [case_names[4]]
-images = range(10)
-# images = [1]
+        itk.transformwrite(
+            [forward_transform],
+            f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_forward.hdf",
+            compression=True,
+        )
 
-input_dir = os.path.join(_HERE, "..", "..", "data", "DirLab-4DCT")
-output_dir = os.path.join(_HERE, "results")
+        itk.transformwrite(
+            [inverse_transform],
+            f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_inverse.hdf",
+            compression=True,
+        )
 
+    # %%
+    seg_image = SegmentChestTotalSegmentator()
+
+    os.makedirs(output_dir, exist_ok=True)
+
+    for case_name in case_names:
+        fixed_image_filename = (
+            f"{input_dir}/{case_name}_T{fixed_image_num * 10:02d}.mhd"
+        )
+        fixed_image = itk.imread(fixed_image_filename)
+        fixed_image = DataDirLab4DCT().fix_image(fixed_image)
+
+        print("Segmenting fixed image...")
+        fixed_result = seg_image.segment(fixed_image)
+        fixed_image_mask = fixed_result["labelmap"]
+        fixed_image_lung_mask = fixed_result["lung"]
+        fixed_image_heart_mask = fixed_result["heart"]
+        fixed_image_major_vessels_mask = fixed_result["major_vessels"]
+        fixed_image_bone_mask = fixed_result["bone"]
+        fixed_image_soft_tissue_mask = fixed_result["soft_tissue"]
+        fixed_image_other_mask = fixed_result["other"]
+        fixed_image_contrast_mask = fixed_result["contrast"]
 
-# %%
-def dilate_mask(mask: Optional[itk.image], dilation: int) -> Optional[itk.image]:
-    if mask is not None:
-        im_math = tube.ImageMath.New(mask)
-        im_math.Dilate(dilation, 1, 0)
-        dilated_mask = im_math.GetOutputShort()
-        return dilated_mask
-    return None
-
-
-def register_image(
-    fixed_image: itk.image,
-    fixed_mask: itk.image,
-    moving_image: itk.image,
-    moving_mask: itk.image,
-    case_name: str,
-    image_num: int,
-    mask_name: str,
-    output_dir: str,
-) -> None:
-    """
-    Register a moving image to a fixed image using a mask.
-    """
-
-    reg_images = RegisterImagesICON()
-    reg_images.set_modality("ct")
-    reg_images.set_number_of_iterations(20)
-
-    if moving_mask is not None:
         itk.imwrite(
-            moving_mask,
-            f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_mask_org.mha",
+            fixed_image_mask,
+            f"{output_dir}/{case_name}_T{fixed_image_num * 10:02d}_mask_org.mha",
             compression=True,
         )
 
-    print("Registering image...")
-    reg_images.set_fixed_image(fixed_image)
-    moving_mask_d = None
-    if fixed_mask is not None:
-        fixed_mask_d = dilate_mask(fixed_mask, heart_mask_dilation)
-        moving_mask_d = dilate_mask(moving_mask, heart_mask_dilation)
-        reg_images.set_fixed_mask(fixed_mask_d)
-    results = reg_images.register(moving_image, moving_mask_d)
-    inverse_transform = results["inverse_transform"]
-    forward_transform = results["forward_transform"]
-    print("Registering image...Done!")
-    moving_image_reg = TransformTools().transform_image(
-        moving_image, forward_transform, fixed_image, "sinc"
-    )  # Final resampling with sinc
-    itk.imwrite(
-        moving_image_reg,
-        f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_reg.mha",
-        compression=True,
-    )
-
-    itk.transformwrite(
-        [forward_transform],
-        f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_forward.hdf",
-        compression=True,
-    )
-
-    itk.transformwrite(
-        [inverse_transform],
-        f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_inverse.hdf",
-        compression=True,
-    )
-
+        # Dynamic anatomy = lung (the structure that moves with respiration)
+        lung_mask_arr = itk.array_from_image(fixed_image_lung_mask)
+        fixed_image_dynamic_anatomy_mask_arr = lung_mask_arr
+        fixed_image_dynamic_anatomy_mask = itk.image_from_array(
+            fixed_image_dynamic_anatomy_mask_arr.astype(np.uint16)
+        )
+        fixed_image_dynamic_anatomy_mask.CopyInformation(fixed_image_mask)
+
+        # Static anatomy = heart, major vessels, contrast, bone, other (all non-lung)
+        heart_mask_arr = itk.array_from_image(fixed_image_heart_mask)
+        major_vessels_mask_arr = itk.array_from_image(fixed_image_major_vessels_mask)
+        contrast_mask_arr = itk.array_from_image(fixed_image_contrast_mask)
+        bone_mask_arr = itk.array_from_image(fixed_image_bone_mask)
+        other_mask_arr = itk.array_from_image(fixed_image_other_mask)
+        fixed_image_static_anatomy_mask_arr = (
+            heart_mask_arr
+            + major_vessels_mask_arr
+            + contrast_mask_arr
+            + bone_mask_arr
+            + other_mask_arr
+        )
+        fixed_image_static_anatomy_mask = itk.image_from_array(
+            fixed_image_static_anatomy_mask_arr.astype(np.uint16)
+        )
+        fixed_image_static_anatomy_mask.CopyInformation(fixed_image_mask)
+        print("Segmenting fixed image...Done!")
 
-# %%
-seg_image = SegmentChestTotalSegmentator()
-
-os.makedirs(output_dir, exist_ok=True)
-
-for case_name in case_names:
-    fixed_image_filename = f"{input_dir}/{case_name}_T{fixed_image_num * 10:02d}.mhd"
-    fixed_image = itk.imread(fixed_image_filename)
-    fixed_image = DataDirLab4DCT().fix_image(fixed_image)
-
-    print("Segmenting fixed image...")
-    fixed_result = seg_image.segment(fixed_image)
-    fixed_image_mask = fixed_result["labelmap"]
-    fixed_image_lung_mask = fixed_result["lung"]
-    fixed_image_heart_mask = fixed_result["heart"]
-    fixed_image_major_vessels_mask = fixed_result["major_vessels"]
-    fixed_image_bone_mask = fixed_result["bone"]
-    fixed_image_soft_tissue_mask = fixed_result["soft_tissue"]
-    fixed_image_other_mask = fixed_result["other"]
-    fixed_image_contrast_mask = fixed_result["contrast"]
-
-    itk.imwrite(
-        fixed_image_mask,
-        f"{output_dir}/{case_name}_T{fixed_image_num * 10:02d}_mask_org.mha",
-        compression=True,
-    )
-
-    # Dynamic anatomy = lung (the structure that moves with respiration)
-    lung_mask_arr = itk.array_from_image(fixed_image_lung_mask)
-    fixed_image_dynamic_anatomy_mask_arr = lung_mask_arr
-    fixed_image_dynamic_anatomy_mask = itk.image_from_array(
-        fixed_image_dynamic_anatomy_mask_arr.astype(np.uint16)
-    )
-    fixed_image_dynamic_anatomy_mask.CopyInformation(fixed_image_mask)
-
-    # Static anatomy = heart, major vessels, contrast, bone, other (all non-lung)
-    heart_mask_arr = itk.array_from_image(fixed_image_heart_mask)
-    major_vessels_mask_arr = itk.array_from_image(fixed_image_major_vessels_mask)
-    contrast_mask_arr = itk.array_from_image(fixed_image_contrast_mask)
-    bone_mask_arr = itk.array_from_image(fixed_image_bone_mask)
-    other_mask_arr = itk.array_from_image(fixed_image_other_mask)
-    fixed_image_static_anatomy_mask_arr = (
-        heart_mask_arr
-        + major_vessels_mask_arr
-        + contrast_mask_arr
-        + bone_mask_arr
-        + other_mask_arr
-    )
-    fixed_image_static_anatomy_mask = itk.image_from_array(
-        fixed_image_static_anatomy_mask_arr.astype(np.uint16)
-    )
-    fixed_image_static_anatomy_mask.CopyInformation(fixed_image_mask)
-    print("Segmenting fixed image...Done!")
-
-    for image_num in images:
-        if image_num != fixed_image_num:
-            moving_image = itk.imread(
-                os.path.join(input_dir, f"{case_name}_T{image_num * 10:02d}.mhd")
-            )
-            moving_image = DataDirLab4DCT().fix_image(moving_image)
-
-            print("***")
-            print("*** Processing case:", case_name, "Image number:", image_num, "***")
-            print("***")
-
-            print("Segmenting moving image...")
-            moving_result = seg_image.segment(moving_image)
-            moving_image_mask = moving_result["labelmap"]
-            moving_image_lung_mask = moving_result["lung"]
-            moving_image_heart_mask = moving_result["heart"]
-            moving_image_major_vessels_mask = moving_result["major_vessels"]
-            moving_image_bone_mask = moving_result["bone"]
-            moving_image_soft_tissue_mask = moving_result["soft_tissue"]
-            moving_image_other_mask = moving_result["other"]
-            moving_image_contrast_mask = moving_result["contrast"]
-
-            # Create heart mask by including major vessels and contrast masks
-            lung_mask_arr = itk.array_from_image(moving_image_lung_mask)
-            moving_image_dynamic_anatomy_mask_arr = lung_mask_arr
-            moving_image_dynamic_anatomy_mask = itk.image_from_array(
-                moving_image_dynamic_anatomy_mask_arr.astype(np.uint16)
-            )
-            moving_image_dynamic_anatomy_mask.CopyInformation(moving_image_mask)
+        for image_num in images:
+            if image_num != fixed_image_num:
+                moving_image = itk.imread(
+                    os.path.join(input_dir, f"{case_name}_T{image_num * 10:02d}.mhd")
+                )
+                moving_image = DataDirLab4DCT().fix_image(moving_image)
 
-            # Create other mask by including lung, bone and soft tissue masks
-            heart_mask_arr = itk.array_from_image(moving_image_heart_mask)
-            major_vessels_mask_arr = itk.array_from_image(
-                moving_image_major_vessels_mask
-            )
-            contrast_mask_arr = itk.array_from_image(moving_image_contrast_mask)
-            bone_mask_arr = itk.array_from_image(moving_image_bone_mask)
-            other_mask_arr = itk.array_from_image(moving_image_other_mask)
-            moving_image_static_anatomy_mask_arr = (
-                heart_mask_arr
-                + major_vessels_mask_arr
-                + contrast_mask_arr
-                + bone_mask_arr
-                + other_mask_arr
-            )
-            moving_image_static_anatomy_mask = itk.image_from_array(
-                moving_image_static_anatomy_mask_arr.astype(np.uint16)
-            )
-            moving_image_static_anatomy_mask.CopyInformation(moving_image_mask)
+                print("***")
+                print(
+                    "*** Processing case:", case_name, "Image number:", image_num, "***"
+                )
+                print("***")
+
+                print("Segmenting moving image...")
+                moving_result = seg_image.segment(moving_image)
+                moving_image_mask = moving_result["labelmap"]
+                moving_image_lung_mask = moving_result["lung"]
+                moving_image_heart_mask = moving_result["heart"]
+                moving_image_major_vessels_mask = moving_result["major_vessels"]
+                moving_image_bone_mask = moving_result["bone"]
+                moving_image_soft_tissue_mask = moving_result["soft_tissue"]
+                moving_image_other_mask = moving_result["other"]
+                moving_image_contrast_mask = moving_result["contrast"]
+
+                # Create heart mask by including major vessels and contrast masks
+                lung_mask_arr = itk.array_from_image(moving_image_lung_mask)
+                moving_image_dynamic_anatomy_mask_arr = lung_mask_arr
+                moving_image_dynamic_anatomy_mask = itk.image_from_array(
+                    moving_image_dynamic_anatomy_mask_arr.astype(np.uint16)
+                )
+                moving_image_dynamic_anatomy_mask.CopyInformation(moving_image_mask)
 
-            print("Segmenting moving image...Done!")
+                # Create other mask by including lung, bone and soft tissue masks
+                heart_mask_arr = itk.array_from_image(moving_image_heart_mask)
+                major_vessels_mask_arr = itk.array_from_image(
+                    moving_image_major_vessels_mask
+                )
+                contrast_mask_arr = itk.array_from_image(moving_image_contrast_mask)
+                bone_mask_arr = itk.array_from_image(moving_image_bone_mask)
+                other_mask_arr = itk.array_from_image(moving_image_other_mask)
+                moving_image_static_anatomy_mask_arr = (
+                    heart_mask_arr
+                    + major_vessels_mask_arr
+                    + contrast_mask_arr
+                    + bone_mask_arr
+                    + other_mask_arr
+                )
+                moving_image_static_anatomy_mask = itk.image_from_array(
+                    moving_image_static_anatomy_mask_arr.astype(np.uint16)
+                )
+                moving_image_static_anatomy_mask.CopyInformation(moving_image_mask)
 
-            itk.imwrite(
-                moving_image_mask,
-                f"{output_dir}/{case_name}_T{image_num * 10:02d}_all_mask_org.mha",
-                compression=True,
-            )
+                print("Segmenting moving image...Done!")
 
-            print("Registering with All mask...")
-            # all
-            register_image(
-                fixed_image,
-                None,
-                moving_image,
-                None,
-                case_name,
-                image_num,
-                "all",
-                output_dir,
-            )
-            print("Registering with All mask...Done!")
-
-            print("Registering with Dynamic Anatomy mask...")
-            # Lungs
-            register_image(
-                fixed_image,
-                fixed_image_dynamic_anatomy_mask,
-                moving_image,
-                moving_image_dynamic_anatomy_mask,
-                case_name,
-                image_num,
-                "dynamic_anatomy",
-                output_dir,
-            )
-            print("Registering with Dynamic Anatomy mask...Done!")
-
-            print("Registering with Static Anatomy mask...")
-            # Bone
-            register_image(
-                fixed_image,
-                fixed_image_static_anatomy_mask,
-                moving_image,
-                moving_image_static_anatomy_mask,
-                case_name,
-                image_num,
-                "static_anatomy",
-                output_dir,
-            )
-            print("Registering with Static Anatomy mask...Done!")
-
-        else:
-            print("Baseline image: no segmentation or registration...")
-            identity_transform = itk.CenteredAffineTransform[itk.D, 3].New()
-            composite_transform = itk.CompositeTransform[itk.D, 3].New()
-            composite_transform.AddTransform(identity_transform)
-
-            for mask, mask_name in [
-                (fixed_image_mask, "all"),
-                (fixed_image_static_anatomy_mask, "static_anatomy"),
-                (fixed_image_dynamic_anatomy_mask, "dynamic_anatomy"),
-            ]:
                 itk.imwrite(
-                    mask,
-                    f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_mask_org.mha",
+                    moving_image_mask,
+                    f"{output_dir}/{case_name}_T{image_num * 10:02d}_all_mask_org.mha",
                     compression=True,
                 )
 
-                itk.imwrite(
+                print("Registering with All mask...")
+                # all
+                register_image(
                     fixed_image,
-                    f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_reg.mha",
-                    compression=True,
+                    None,
+                    moving_image,
+                    None,
+                    case_name,
+                    image_num,
+                    "all",
+                    output_dir,
                 )
+                print("Registering with All mask...Done!")
 
-                itk.transformwrite(
-                    [composite_transform],
-                    f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_forward.hdf",
-                    compression=True,
+                print("Registering with Dynamic Anatomy mask...")
+                # Lungs
+                register_image(
+                    fixed_image,
+                    fixed_image_dynamic_anatomy_mask,
+                    moving_image,
+                    moving_image_dynamic_anatomy_mask,
+                    case_name,
+                    image_num,
+                    "dynamic_anatomy",
+                    output_dir,
                 )
+                print("Registering with Dynamic Anatomy mask...Done!")
 
-                itk.transformwrite(
-                    [composite_transform],
-                    f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_inverse.hdf",
-                    compression=True,
+                print("Registering with Static Anatomy mask...")
+                # Bone
+                register_image(
+                    fixed_image,
+                    fixed_image_static_anatomy_mask,
+                    moving_image,
+                    moving_image_static_anatomy_mask,
+                    case_name,
+                    image_num,
+                    "static_anatomy",
+                    output_dir,
                 )
-
-            print("Baseline image: no segmentation or registration...Done!")
+                print("Registering with Static Anatomy mask...Done!")
+
+            else:
+                print("Baseline image: no segmentation or registration...")
+                identity_transform = itk.CenteredAffineTransform[itk.D, 3].New()
+                composite_transform = itk.CompositeTransform[itk.D, 3].New()
+                composite_transform.AddTransform(identity_transform)
+
+                for mask, mask_name in [
+                    (fixed_image_mask, "all"),
+                    (fixed_image_static_anatomy_mask, "static_anatomy"),
+                    (fixed_image_dynamic_anatomy_mask, "dynamic_anatomy"),
+                ]:
+                    itk.imwrite(
+                        mask,
+                        f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_mask_org.mha",
+                        compression=True,
+                    )
+
+                    itk.imwrite(
+                        fixed_image,
+                        f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_reg.mha",
+                        compression=True,
+                    )
+
+                    itk.transformwrite(
+                        [composite_transform],
+                        f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_forward.hdf",
+                        compression=True,
+                    )
+
+                    itk.transformwrite(
+                        [composite_transform],
+                        f"{output_dir}/{case_name}_T{image_num * 10:02d}_{mask_name}_inverse.hdf",
+                        compression=True,
+                    )
+
+                print("Baseline image: no segmentation or registration...Done!")
diff --git a/experiments/Lung-GatedCT_To_USD/1-make_dirlab_models.py b/experiments/Lung-GatedCT_To_USD/1-make_dirlab_models.py
index f8d7e60..252ce0a 100644
--- a/experiments/Lung-GatedCT_To_USD/1-make_dirlab_models.py
+++ b/experiments/Lung-GatedCT_To_USD/1-make_dirlab_models.py
@@ -9,104 +9,112 @@
 from physiomotion4d import ConvertVTKToUSD
 from physiomotion4d.segment_chest_total_segmentator import SegmentChestTotalSegmentator
 
+# Defensive: today this script only reads `seg.all_mask_ids`, but if anyone
+# adds a `seg.segment(...)` call it would trigger the nnUNet
+# multiprocessing.Pool which re-imports the script on Windows (spawn start
+# method) and crashes with a spawn-cascade RuntimeError. Guard pre-emptively.
+if __name__ == "__main__":
+    case_names = DataDirLab4DCT().case_names
+    case_names = [case_names[4]]
+
+    base_timepoint = 30
+
+    output_dir = "./results"
+
+    # %%
+    def transform_contours_list(
+        contours: pv.PolyData, case_name: str, mask_name: str, output_dir: str
+    ):
+        """
+        Transform a list of contours to a list of transformed contours.
+        """
+        con_tools = ContourTools()
+        new_contours = []
+        for i in range(10):
+            inverse_transform = itk.transformread(
+                f"{output_dir}/{case_name}_T{i * 10:02d}_{mask_name}_inverse.hdf"
+            )[0]
+
+            print(f"Transforming {case_name} - {mask_name} - T{i * 10:02d}")
+            new_contours.append(
+                con_tools.transform_contours(contours, inverse_transform)
+            )
+
+        return new_contours
+
+    # %%
+    def make_dirlab_models(
+        output_dir,
+        label,
+        case_name,
+        base_timepoint,
+        all_labelmap_arr,
+        all_mask_ids,
+        con_tools,
+        seg,
+    ):
+        """
+        Make DirLab models for a list of cases.
+        """
+        labelmap_image = itk.imread(
+            f"{output_dir}/{case_name}_T{base_timepoint}_{label}_mask_org.mha",
+            pixel_type=itk.UC,
+        )
+        labelmap_arr = itk.array_view_from_image(labelmap_image)
 
-case_names = DataDirLab4DCT().case_names
-case_names = [case_names[4]]
+        print(f"Extracting contours from {case_name} - {label} Contours")
+        label_labelmap_arr = np.where(labelmap_arr > 0, all_labelmap_arr, 0).astype(
+            np.uint8
+        )
+        label_labelmap_image = itk.image_from_array(label_labelmap_arr)
+        label_labelmap_image.CopyInformation(labelmap_image)
 
-base_timepoint = 30
+        contours = con_tools.extract_contours(label_labelmap_image)
+        contours.save(
+            f"{output_dir}/{case_name}_T{base_timepoint}_{label}_lungGatedBase.vtp",
+            binary=True,
+        )
 
-output_dir = "./results"
+        print(f"Applying transforms to vtp models from {case_name}")
+        transformed_contours = transform_contours_list(
+            contours, case_name, label, output_dir
+        )
 
+        print(f"Converting vtp models to USD for {case_name}")
+        # Forwarding `seg` groups labels by anatomy type under
+        # /World/DirLab4DCT/{type}/{label_name}.
+        converter = ConvertVTKToUSD(
+            "DirLab4DCT",
+            transformed_contours,
+            mask_ids=all_mask_ids,
+            segmenter=seg,
+        )
+        converter.convert(
+            f"{output_dir}/{case_name}_{label}_lungGated.usd",
+            convert_to_surface=True,
+        )
 
-# %%
-def transform_contours_list(
-    contours: pv.PolyData, case_name: str, mask_name: str, output_dir: str
-):
-    """
-    Transform a list of contours to a list of transformed contours.
-    """
+    # %%
     con_tools = ContourTools()
-    new_contours = []
-    for i in range(10):
-        inverse_transform = itk.transformread(
-            f"{output_dir}/{case_name}_T{i * 10:02d}_{mask_name}_inverse.hdf"
-        )[0]
-
-        print(f"Transforming {case_name} - {mask_name} - T{i * 10:02d}")
-        new_contours.append(con_tools.transform_contours(contours, inverse_transform))
-
-    return new_contours
-
-
-# %%
-def make_dirlab_models(
-    output_dir,
-    label,
-    case_name,
-    base_timepoint,
-    all_labelmap_arr,
-    all_mask_ids,
-    con_tools,
-):
-    """
-    Make DirLab models for a list of cases.
-    """
-    labelmap_image = itk.imread(
-        f"{output_dir}/{case_name}_T{base_timepoint}_{label}_mask_org.mha",
-        pixel_type=itk.UC,
-    )
-    labelmap_arr = itk.array_view_from_image(labelmap_image)
-
-    print(f"Extracting contours from {case_name} - {label} Contours")
-    label_labelmap_arr = np.where(labelmap_arr > 0, all_labelmap_arr, 0).astype(
-        np.uint8
-    )
-    label_labelmap_image = itk.image_from_array(label_labelmap_arr)
-    label_labelmap_image.CopyInformation(labelmap_image)
-
-    contours = con_tools.extract_contours(label_labelmap_image)
-    contours.save(
-        f"{output_dir}/{case_name}_T{base_timepoint}_{label}_lungGatedBase.vtp",
-        binary=True,
-    )
-
-    print(f"Applying transforms to vtp models from {case_name}")
-    transformed_contours = transform_contours_list(
-        contours, case_name, label, output_dir
-    )
-
-    print(f"Converting vtp models to USD for {case_name}")
-    converter = ConvertVTKToUSD(
-        "DirLab4DCT",
-        transformed_contours,
-        mask_ids=all_mask_ids,
-    )
-    converter.convert(
-        f"{output_dir}/{case_name}_{label}_lungGated.usd",
-        convert_to_surface=True,
-    )
-
-
-# %%
-con_tools = ContourTools()
-
-seg = SegmentChestTotalSegmentator()
-for case_name in case_names:
-    # all labelmap
-    all_labelmap = itk.imread(
-        f"{output_dir}/{case_name}_T{base_timepoint}_all_mask_org.mha",
-        pixel_type=itk.UC,
-    )
-    all_labelmap_arr = itk.array_view_from_image(all_labelmap)
-    all_mask_ids = seg.all_mask_ids
 
-    for label in ["all", "static_anatomy", "dynamic_anatomy"]:
-        make_dirlab_models(
-            output_dir,
-            label,
-            case_name,
-            base_timepoint,
-            all_labelmap_arr,
-            all_mask_ids,
-            con_tools,
+    seg = SegmentChestTotalSegmentator()
+    for case_name in case_names:
+        # all labelmap
+        all_labelmap = itk.imread(
+            f"{output_dir}/{case_name}_T{base_timepoint}_all_mask_org.mha",
+            pixel_type=itk.UC,
         )
+        all_labelmap_arr = itk.array_view_from_image(all_labelmap)
+        all_mask_ids = seg.taxonomy.all_labels()
+
+        for label in ["all", "static_anatomy", "dynamic_anatomy"]:
+            make_dirlab_models(
+                output_dir,
+                label,
+                case_name,
+                base_timepoint,
+                all_labelmap_arr,
+                all_mask_ids,
+                con_tools,
+                seg,
+            )
diff --git a/experiments/Lung-GatedCT_To_USD/2-paint_dirlab_models.py b/experiments/Lung-GatedCT_To_USD/2-paint_dirlab_models.py
index c101275..dc1dd71 100644
--- a/experiments/Lung-GatedCT_To_USD/2-paint_dirlab_models.py
+++ b/experiments/Lung-GatedCT_To_USD/2-paint_dirlab_models.py
@@ -6,19 +6,24 @@
 from physiomotion4d.segment_chest_total_segmentator import SegmentChestTotalSegmentator
 from physiomotion4d.usd_anatomy_tools import USDAnatomyTools
 
+# Defensive: today this script only instantiates SegmentChestTotalSegmentator
+# to read its anatomy labels for USDAnatomyTools, but if anyone adds a
+# `seg.segment(...)` call it would trigger the nnUNet multiprocessing.Pool
+# which re-imports the script on Windows (spawn start method) and crashes
+# with a spawn-cascade RuntimeError. Guard pre-emptively.
+if __name__ == "__main__":
+    case_names = DataDirLab4DCT().case_names
 
-case_names = DataDirLab4DCT().case_names
+    case_names = [case_names[4]]
 
-case_names = [case_names[4]]
+    output_dir = "./results"
 
-output_dir = "./results"
+    # %%
+    seg = SegmentChestTotalSegmentator()
 
-# %%
-seg = SegmentChestTotalSegmentator()
-
-for anatomy in ["all", "static_anatomy", "dynamic_anatomy"]:
-    for case_name in case_names:
-        stage = Usd.Stage.Open(f"{output_dir}/{case_name}_{anatomy}_lungGated.usd")
-        painter = USDAnatomyTools(stage)
-        painter.enhance_meshes(seg)
-        stage.Export(f"{output_dir}/{case_name}_{anatomy}_lungGated_painted.usd")
+    for anatomy in ["all", "static_anatomy", "dynamic_anatomy"]:
+        for case_name in case_names:
+            stage = Usd.Stage.Open(f"{output_dir}/{case_name}_{anatomy}_lungGated.usd")
+            painter = USDAnatomyTools(stage)
+            painter.enhance_meshes(seg)
+            stage.Export(f"{output_dir}/{case_name}_{anatomy}_lungGated_painted.usd")
diff --git a/experiments/Lung-GatedCT_To_USD/Experiment_SegReg.py b/experiments/Lung-GatedCT_To_USD/Experiment_SegReg.py
index a428df7..54fa954 100644
--- a/experiments/Lung-GatedCT_To_USD/Experiment_SegReg.py
+++ b/experiments/Lung-GatedCT_To_USD/Experiment_SegReg.py
@@ -9,37 +9,42 @@
 from physiomotion4d import RegisterImagesICON
 from physiomotion4d import SegmentChestTotalSegmentator
 
-_HERE = os.path.dirname(os.path.abspath(__file__))
-_DATA_DIR = os.path.join(_HERE, "..", "..", "data", "DirLab-4DCT")
-_RESULTS_DIR = os.path.join(_HERE, "results_SegReg")
+# nnUNetv2 (used by TotalSegmentator) spawns a multiprocessing.Pool. On Windows
+# the spawn start method re-imports this script in each child; without the
+# __name__ == "__main__" guard around the top-level work, that re-import fires
+# segment() again and Python's spawn-cascade detector raises RuntimeError.
+if __name__ == "__main__":
+    _HERE = os.path.dirname(os.path.abspath(__file__))
+    _DATA_DIR = os.path.join(_HERE, "..", "..", "data", "DirLab-4DCT")
+    _RESULTS_DIR = os.path.join(_HERE, "results_SegReg")
 
-# %%
-fixed_image = DataDirLab4DCT().fix_image(
-    itk.imread(os.path.join(_DATA_DIR, "Case1Pack_T30.mhd"))
-)
-moving_image = DataDirLab4DCT().fix_image(
-    itk.imread(os.path.join(_DATA_DIR, "Case1Pack_T00.mhd"))
-)
+    # %%
+    fixed_image = DataDirLab4DCT().fix_image(
+        itk.imread(os.path.join(_DATA_DIR, "Case1Pack_T30.mhd"))
+    )
+    moving_image = DataDirLab4DCT().fix_image(
+        itk.imread(os.path.join(_DATA_DIR, "Case1Pack_T00.mhd"))
+    )
 
-# %%
-# Register images
-reg_images = RegisterImagesICON()
-reg_images.set_fixed_image(fixed_image)
-_ = reg_images.register(moving_image)
-moving_image_registered = reg_images.get_registered_image()
-os.makedirs(_RESULTS_DIR, exist_ok=True)
-itk.imwrite(
-    moving_image_registered,
-    os.path.join(_RESULTS_DIR, "Experiment_reg.mha"),
-    compression=True,
-)
+    # %%
+    # Register images
+    reg_images = RegisterImagesICON()
+    reg_images.set_fixed_image(fixed_image)
+    _ = reg_images.register(moving_image)
+    moving_image_registered = reg_images.get_registered_image()
+    os.makedirs(_RESULTS_DIR, exist_ok=True)
+    itk.imwrite(
+        moving_image_registered,
+        os.path.join(_RESULTS_DIR, "Experiment_reg.mha"),
+        compression=True,
+    )
 
-# %%
-img = itk.imread(os.path.join(_RESULTS_DIR, "Experiment_reg.mha"))
-tot_seg = SegmentChestTotalSegmentator()
-seg_results = tot_seg.segment(img, contrast_enhanced_study=False)
-itk.imwrite(
-    seg_results["labelmap"],
-    os.path.join(_RESULTS_DIR, "Experiment_totseg.mha"),
-    compression=True,
-)
+    # %%
+    img = itk.imread(os.path.join(_RESULTS_DIR, "Experiment_reg.mha"))
+    tot_seg = SegmentChestTotalSegmentator()
+    seg_results = tot_seg.segment(img, contrast_enhanced_study=False)
+    itk.imwrite(
+        seg_results["labelmap"],
+        os.path.join(_RESULTS_DIR, "Experiment_totseg.mha"),
+        compression=True,
+    )
diff --git a/experiments/README.md b/experiments/README.md
index ebf5e64..7f975a6 100644
--- a/experiments/README.md
+++ b/experiments/README.md
@@ -131,7 +131,7 @@ model from a population of anatomical meshes using the KCL Heart Model dataset a
 - Disease-specific shape analysis (pathology vs. normal)
 - Surgical planning with shape prediction
 
-**⚠️ Required first**: Complete this experiment BEFORE `Heart-Statistical_Model_To_Patient`
+**Required first**: Complete this experiment BEFORE `Heart-Statistical_Model_To_Patient`
 as it generates the PCA model files needed for patient-specific registration.
 
 **Output:** PCA model files (eigenvectors, eigenvalues, mean shape, template mask) for
@@ -211,11 +211,11 @@ pytest tests/test_experiments.py::test_list_scripts_in_subdir -v -s --run-experi
 pytest tests/test_experiments.py::test_experiment_structure -v --run-experiments
 ```
 
-🔒 **IMPORTANT:** Experiment tests require the `--run-experiments` flag. Without this flag, they are automatically skipped, even if you run `pytest tests/` or target the test file directly.
+**IMPORTANT:** Experiment tests require the `--run-experiments` flag. Without this flag, they are automatically skipped, even if you run `pytest tests/` or target the test file directly.
 
 ### Test Features
 
-- **Test-mode flag** - When run as tests (pytest with `--run-experiments`), the runner sets `PHYSIOMOTION_RUNNING_AS_TEST=1`. Scripts can read this (e.g. via `physiomotion4d.notebook_utils.running_as_test()`) and use reduced parameters so test runs stay fast. See [tests/EXPERIMENT_TESTS_GUIDE.md](../tests/EXPERIMENT_TESTS_GUIDE.md#running-as-test-physiomotion_running_as_test).
+- **Test-mode flag** - When run as tests (pytest with `--run-experiments`), the runner sets `PHYSIOMOTION_RUNNING_AS_TEST=1`. Scripts can read this (e.g. via `physiomotion4d.test_tools.TestTools.running_as_test()`) and use reduced parameters so test runs stay fast. See [tests/EXPERIMENT_TESTS_GUIDE.md](../tests/EXPERIMENT_TESTS_GUIDE.md#running-as-test-physiomotion_running_as_test).
 - **One test per subdirectory** - Each experiment subdirectory gets its own test function
 - **Alphanumeric ordering** - Scripts execute in alphanumeric order (e.g., `0-`, `1-`, `2-`)
 - **Long timeouts** - Each script has up to 1 hour execution time, tests have multi-hour timeouts
@@ -233,11 +233,11 @@ These tests require:
 
 ### Important Notes
 
-⚠️ **These tests are extremely long-running** - Plan for multiple hours of execution time
+**These tests are extremely long-running** - Plan for multiple hours of execution time
 
-⚠️ **Not part of CI/CD** - These tests are excluded from all automated workflows
+**Not part of CI/CD** - These tests are excluded from all automated workflows
 
-⚠️ **Resource intensive** - Requires GPU, significant memory, and disk space
+**Resource intensive** - Requires GPU, significant memory, and disk space
 
 ## Structure
 
@@ -262,16 +262,16 @@ Each subdirectory represents a different experimental domain:
 
 ## Important Notes
 
-⚠️ **These experiments are not production code** - They may contain:
+**These experiments are not production code** - They may contain:
 - Hardcoded paths and parameters
 - Incomplete error handling
 - Experimental APIs that have since changed
 - Code that doesn't follow the final library conventions
 - Dataset-specific assumptions and configurations
 
-⚠️ **Do not use these as usage examples** - These are research artifacts, not tutorials.
+**Do not use these as usage examples** - These are research artifacts, not tutorials.
 
-### ✅ For Production Use, Consult:
+### For Production Use, Consult:
 
 1. **CLI Commands** ⭐ **PRIMARY RESOURCE**
    - `physiomotion4d-heart-gated-ct` - Complete heart-gated CT workflow
diff --git a/experiments/Reconstruct4DCT/reconstruct_4d_ct_class.py b/experiments/Reconstruct4DCT/reconstruct_4d_ct_class.py
index 184015d..43fd937 100644
--- a/experiments/Reconstruct4DCT/reconstruct_4d_ct_class.py
+++ b/experiments/Reconstruct4DCT/reconstruct_4d_ct_class.py
@@ -17,7 +17,7 @@
 import numpy as np
 
 from physiomotion4d import RegisterTimeSeriesImages, TransformTools
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 
 _HERE = os.path.dirname(os.path.abspath(__file__))
 
@@ -40,7 +40,7 @@
 
 # %%
 # Configuration: quick run when executed as test (pytest); full run when manual (set quick_run = True for interactive quick test)
-quick_run = running_as_test()
+quick_run = TestTools.running_as_test()
 
 # Select files and parameters based on mode
 if quick_run:
diff --git a/pyproject.toml b/pyproject.toml
index b1f3b23..d318d42 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -14,7 +14,7 @@ maintainers = [
 ]
 readme = {file = "README.md", content-type = "text/markdown"}
 license = {text = "Apache-2.0"}
-requires-python = ">=3.10"
+requires-python = ">=3.11"
 classifiers = [
     "Development Status :: 4 - Beta",
     "Intended Audience :: Healthcare Industry",
@@ -22,7 +22,6 @@ classifiers = [
     "License :: OSI Approved :: Apache Software License",
     "Operating System :: OS Independent",
     "Programming Language :: Python :: 3",
-    "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",
     "Programming Language :: Python :: 3.12",
     "Topic :: Scientific/Engineering :: Medical Science Apps.",
@@ -63,6 +62,7 @@ dependencies = [
 
     # AI/ML and segmentation
     "monai>=1.3.0",
+    "nvidia-physicsnemo>=2.0.0,<3.0.0",
     "torch>=2.0.0,<3.0.0",
     "transformers>=4.21.0,<5.0.0",
     "totalsegmentator>=2.0.0",
@@ -75,11 +75,12 @@ dependencies = [
     "unigradicon>=1.0.0",
 
     # Visualization and USD
-    "pyvista[all]>=0.43.0",
+    "pyvista[all]>=0.47.0",
     "usd-core>=23.11",
     "trimesh>=4.0.0",
 
     # Utilities
+    "ipykernel>=6.0.0",
     "matplotlib>=3.5.0",
     "typing-extensions>=4.0.0",
 
@@ -188,7 +189,7 @@ push = false
 ]
 
 [tool.mypy]
-python_version = "3.10"
+python_version = "3.11"
 warn_return_any = true
 warn_unused_configs = true
 disallow_untyped_defs = true
diff --git a/src/physiomotion4d/__init__.py b/src/physiomotion4d/__init__.py
index 9092580..2fed426 100644
--- a/src/physiomotion4d/__init__.py
+++ b/src/physiomotion4d/__init__.py
@@ -33,11 +33,13 @@
         stacklevel=2,
     )
 
+from .anatomy_taxonomy import AnatomyGroup, AnatomyTaxonomy
 from .contour_tools import ContourTools
 
 # Data processing utilities
 from .convert_nrrd_4d_to_3d import ConvertNRRD4DTo3D
 from .convert_vtk_to_usd import ConvertVTKToUSD
+from .data_download_tools import DataDownloadTools
 
 # Utility classes
 from .image_tools import ImageTools
@@ -106,7 +108,11 @@
     "USDTools",
     "ContourTools",
     "USDAnatomyTools",
+    "DataDownloadTools",
     # Data processing utilities
     "ConvertNRRD4DTo3D",
     "ConvertVTKToUSD",
+    # Anatomy taxonomy (shared between segmenters and USD renderer)
+    "AnatomyTaxonomy",
+    "AnatomyGroup",
 ]
diff --git a/src/physiomotion4d/anatomy_taxonomy.py b/src/physiomotion4d/anatomy_taxonomy.py
new file mode 100644
index 0000000..55b1c9c
--- /dev/null
+++ b/src/physiomotion4d/anatomy_taxonomy.py
@@ -0,0 +1,164 @@
+"""Pure-data anatomy taxonomy shared by segmenters and USD renderers.
+
+This module defines :class:`AnatomyTaxonomy` and :class:`AnatomyGroup`, a
+minimal data type that maps anatomical groups (``heart``, ``lung``, ``bone``,
+...) to the organ labels they contain.
+
+The taxonomy is the single source of truth for the label hierarchy. Both
+:class:`physiomotion4d.SegmentAnatomyBase` (which populates one via its
+subclasses) and :class:`physiomotion4d.USDAnatomyTools` (which consumes one
+when applying materials) depend on this class. The two consumers do not
+depend on each other, which lets either side be used without the other.
+
+The class is pure stdlib (`dataclasses`, `typing`); it does not import
+``itk``, ``pxr``, or any other heavy dependency.
+"""
+
+from __future__ import annotations
+
+import logging
+from dataclasses import dataclass, field
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class AnatomyGroup:
+    """One named anatomy group together with the organ labels it contains.
+
+    Attributes:
+        name: Group name (e.g. ``"heart"``, ``"lung"``).
+        organs: Maps integer label id to organ name within this group.
+    """
+
+    name: str
+    organs: dict[int, str] = field(default_factory=dict)
+
+
+class AnatomyTaxonomy:
+    """Mapping of anatomical groups to the organs each group contains.
+
+    Groups are added in insertion order, which determines the order returned
+    by :meth:`group_names` and the iteration order of :meth:`all_labels`.
+
+    Example:
+        >>> tax = AnatomyTaxonomy()
+        >>> tax.add_organ("heart", 51, "heart")
+        >>> tax.add_organ("heart", 61, "atrial_appendage_left")
+        >>> tax.add_organ("lung", 10, "lung_upper_lobe_left")
+        >>> tax.group_for_label("atrial_appendage_left")
+        'heart'
+        >>> tax.labels_in_group("lung")
+        {10: 'lung_upper_lobe_left'}
+    """
+
+    OTHER_GROUP = "other"
+    """Sentinel group name used by :meth:`fill_other_group` for unclaimed ids."""
+
+    def __init__(self) -> None:
+        self._groups: dict[str, AnatomyGroup] = {}
+
+    def add_group(self, name: str) -> AnatomyGroup:
+        """Ensure a group exists and return it.
+
+        Args:
+            name: Group name.
+
+        Returns:
+            The (new or existing) :class:`AnatomyGroup`.
+        """
+        if name not in self._groups:
+            self._groups[name] = AnatomyGroup(name=name)
+        return self._groups[name]
+
+    def add_organ(self, group: str, label_id: int, organ_name: str) -> None:
+        """Add one organ label to the named group.
+
+        Creates the group if it does not yet exist. Reassigning the same
+        label id within the same group is silently allowed (last write wins).
+        If the same id is already registered in a *different* group, a warning
+        is logged and the new assignment is dropped, so the first registration
+        wins and :meth:`group_for_id` remains deterministic.
+
+        Args:
+            group: Target group name.
+            label_id: Integer label id (e.g. TotalSegmentator class index).
+            organ_name: Human-readable organ name.
+        """
+        for existing in self._groups.values():
+            if existing.name == group:
+                continue
+            if label_id in existing.organs:
+                logger.warning(
+                    "label_id %d already registered in group %r as %r; "
+                    "ignoring duplicate add to group %r as %r",
+                    label_id,
+                    existing.name,
+                    existing.organs[label_id],
+                    group,
+                    organ_name,
+                )
+                return
+        self.add_group(group).organs[label_id] = organ_name
+
+    def group_names(self) -> list[str]:
+        """Return group names in the order they were first added."""
+        return list(self._groups.keys())
+
+    def labels_in_group(self, group: str) -> dict[int, str]:
+        """Return ``{label_id: organ_name}`` for *group*; empty dict if absent."""
+        anatomy_group = self._groups.get(group)
+        return dict(anatomy_group.organs) if anatomy_group is not None else {}
+
+    def all_labels(self) -> dict[int, str]:
+        """Return the union of every group's organs as a single id→name dict."""
+        merged: dict[int, str] = {}
+        for anatomy_group in self._groups.values():
+            merged.update(anatomy_group.organs)
+        return merged
+
+    def group_for_label(self, label_name: str) -> str:
+        """Return the group containing *label_name*.
+
+        Falls back to :data:`OTHER_GROUP` if no group contains the name; this
+        keeps :class:`physiomotion4d.ConvertVTKToUSD` happy when it encounters
+        labels the segmenter did not classify.
+        """
+        for anatomy_group in self._groups.values():
+            if label_name in anatomy_group.organs.values():
+                return anatomy_group.name
+        return self.OTHER_GROUP
+
+    def group_for_id(self, label_id: int) -> str:
+        """Return the group containing *label_id*; :data:`OTHER_GROUP` if absent."""
+        for anatomy_group in self._groups.values():
+            if label_id in anatomy_group.organs:
+                return anatomy_group.name
+        return self.OTHER_GROUP
+
+    def fill_other_group(
+        self,
+        id_range: range = range(1, 256),
+        name_template: str = "other_{id}",
+    ) -> None:
+        """Populate the ``other`` group with any ids not already claimed.
+
+        Called by :class:`physiomotion4d.SegmentAnatomyBase` subclasses at the
+        end of ``__init__`` to mark every id in the segmenter's class index
+        space that no specific group claimed.
+
+        Args:
+            id_range: Inclusive-lower / exclusive-upper id space to scan.
+            name_template: ``str.format`` template for synthetic organ names;
+                receives the unclaimed id as ``{id}``.
+        """
+        claimed_ids: set[int] = set()
+        for anatomy_group in self._groups.values():
+            if anatomy_group.name == self.OTHER_GROUP:
+                continue
+            claimed_ids.update(anatomy_group.organs.keys())
+        for label_id in id_range:
+            if label_id not in claimed_ids:
+                self.add_organ(
+                    self.OTHER_GROUP, label_id, name_template.format(id=label_id)
+                )
diff --git a/src/physiomotion4d/cli/convert_vtk_to_usd.py b/src/physiomotion4d/cli/convert_vtk_to_usd.py
index 66549a1..1796627 100644
--- a/src/physiomotion4d/cli/convert_vtk_to_usd.py
+++ b/src/physiomotion4d/cli/convert_vtk_to_usd.py
@@ -11,18 +11,12 @@
 import os
 import sys
 
-ANATOMY_TYPES = [
-    "heart",
-    "lung",
-    "bone",
-    "major_vessels",
-    "contrast",
-    "soft_tissue",
-    "other",
-    "liver",
-    "spleen",
-    "kidney",
-]
+from physiomotion4d.usd_anatomy_tools import DEFAULT_RENDER_PARAMS
+
+# Anatomy types accepted by --anatomy-type, sourced from the renderer's
+# registered defaults so that new groups/organs registered in
+# DEFAULT_RENDER_PARAMS automatically appear here without code changes.
+ANATOMY_TYPES = list(DEFAULT_RENDER_PARAMS.keys())
 
 
 def main() -> int:
diff --git a/src/physiomotion4d/convert_nrrd_4d_to_3d.py b/src/physiomotion4d/convert_nrrd_4d_to_3d.py
index 069e8c9..9ce75ae 100644
--- a/src/physiomotion4d/convert_nrrd_4d_to_3d.py
+++ b/src/physiomotion4d/convert_nrrd_4d_to_3d.py
@@ -1,10 +1,12 @@
 import logging
-from typing import Any, Optional
+from pathlib import Path
+from typing import Any, Optional, Union
 
 import itk
 import nrrd
 import numpy as np
 
+
 from physiomotion4d.physiomotion4d_base import PhysioMotion4DBase
 
 
@@ -65,6 +67,12 @@ def get_3d_image(self, index: int) -> Any:
     def get_number_of_3d_images(self) -> int:
         return len(self.img_3d)
 
-    def save_3d_images(self, basename: str) -> None:
+    def save_3d_images(self, directory: Union[str, Path], basename: str) -> None:
+        dir_path = Path(directory)
+        dir_path.mkdir(parents=True, exist_ok=True)
         for i in range(self.get_number_of_3d_images()):
-            itk.imwrite(self.img_3d[i], f"{basename}_{i:03d}.mha", compression=True)
+            itk.imwrite(
+                self.img_3d[i],
+                str(dir_path / f"{basename}_{i:03d}.mha"),
+                compression=True,
+            )
diff --git a/src/physiomotion4d/convert_vtk_to_usd.py b/src/physiomotion4d/convert_vtk_to_usd.py
index 7ad8365..8c1b821 100644
--- a/src/physiomotion4d/convert_vtk_to_usd.py
+++ b/src/physiomotion4d/convert_vtk_to_usd.py
@@ -21,9 +21,10 @@
 import numpy as np
 import pyvista as pv
 import vtk
-from pxr import Usd, UsdGeom
+from pxr import Sdf, Usd, UsdGeom
 
 from .physiomotion4d_base import PhysioMotion4DBase
+from .segment_anatomy_base import SegmentAnatomyBase
 from .vtk_to_usd import (
     ConversionSettings,
     DataType,
@@ -32,6 +33,7 @@
     MaterialManager,
     MeshData,
     UsdMeshConverter,
+    add_framing_camera,
     cell_type_name_for_vertex_count,
     read_vtk_file,
     split_mesh_data_by_cell_type,
@@ -80,6 +82,7 @@ def __init__(
         times_per_second: float = 24.0,
         separate_by: Literal["none", "connectivity", "cell_type"] = "none",
         solid_color: tuple[float, float, float] = (0.8, 0.8, 0.8),
+        segmenter: Optional[SegmentAnatomyBase] = None,
         log_level: int | str = logging.INFO,
     ) -> None:
         """
@@ -98,6 +101,12 @@ def __init__(
                         'none' keeps the mesh as-is, 'connectivity' splits by connected
                         component, 'cell_type' splits by face vertex count.
             solid_color: Default RGB diffuse color in [0, 1] used when no colormap is set.
+            segmenter: Optional SegmentAnatomyBase instance used to classify each
+                       mask_ids label into an anatomy group (heart / lung / bone /
+                       major_vessels / contrast / soft_tissue / other) so labeled
+                       prims are written under ``/World/{data_basename}/{type}/{label}``
+                       and materials under ``/World/Looks/{type}/{label}_material``.
+                       When None, labels are grouped under a single ``Anatomy`` Xform.
             log_level: Logging level
         """
         super().__init__(class_name=self.__class__.__name__, log_level=log_level)
@@ -109,6 +118,7 @@ def __init__(
         self.convert_to_surface = convert_to_surface
         self.separate_by = separate_by
         self.solid_color = solid_color
+        self.segmenter = segmenter
 
         # Colormap settings
         self.color_by_array: Optional[str] = None
@@ -153,6 +163,7 @@ def from_files(
         time_codes: Optional[list[float]] = None,
         static_merge: bool = False,
         mask_ids: Optional[dict[int, str]] = None,
+        segmenter: Optional[SegmentAnatomyBase] = None,
         log_level: int | str = logging.INFO,
     ) -> "ConvertVTKToUSD":
         """Create a converter by loading VTK files from disk.
@@ -173,6 +184,8 @@ def from_files(
             static_merge: If True, treat each file as a separate mesh object in a
                           single static scene (no time samples).
             mask_ids: Optional anatomical label mapping.
+            segmenter: Optional SegmentAnatomyBase used to group labeled prims
+                       by anatomy type. See the ConvertVTKToUSD constructor.
             log_level: Logging level.
 
         Returns:
@@ -217,6 +230,7 @@ def from_files(
             convert_to_surface=extract_surface,
             times_per_second=times_per_second,
             solid_color=solid_color,
+            segmenter=segmenter,
             log_level=log_level,
         )
         instance._is_static_merge = static_merge
@@ -420,6 +434,107 @@ def set_colormap(
 
         return self
 
+    def compute_von_mises_stress(
+        self,
+        stress_array_name: str = "stress",
+        output_name: str = "von_mises_stress",
+    ) -> "ConvertVTKToUSD":
+        """Add a scalar von Mises stress array derived from a 9-component
+        stress tensor on every input mesh.
+
+        For each mesh in ``self.input_polydata``, looks up ``stress_array_name``
+        in ``cell_data`` first (FE stress is typically per-cell), then in
+        ``point_data``, reduces the 9-component tensor to scalar von Mises, and
+        writes the result back to the same data dict under ``output_name``.
+
+        Call this between ``from_files()`` and ``convert()``. The new array
+        becomes a USD primvar at convert time (``vtk_cell_<output_name>`` or
+        ``vtk_point_<output_name>``) and can be selected as the
+        ``color_by_array`` for set_colormap or as the target primvar for
+        ``USDTools.apply_colormap_from_primvar``.
+
+        Tensor layout (row-major)::
+
+            [s_xx, s_xy, s_xz, s_yx, s_yy, s_yz, s_zx, s_zy, s_zz]
+
+        Off-diagonal pairs are averaged to symmetrize, which is a no-op for an
+        already-symmetric Cauchy stress tensor.
+
+        Formula::
+
+            sigma_VM = sqrt(0.5 * [(sxx-syy)^2 + (syy-szz)^2 + (szz-sxx)^2]
+                            + 3.0 * (sxy^2 + syz^2 + szx^2))
+
+        Args:
+            stress_array_name: Source array name on the input meshes.
+                Defaults to ``"stress"``.
+            output_name: Name under which to store the resulting scalar
+                array. Defaults to ``"von_mises_stress"``.
+
+        Returns:
+            self: For method chaining.
+
+        Raises:
+            ValueError: If no input mesh contains a 9-component array named
+                ``stress_array_name`` in either cell_data or point_data.
+        """
+        found_any = False
+        for mesh in self.input_polydata:
+            pv_mesh = mesh if isinstance(mesh, pv.DataSet) else pv.wrap(mesh)
+            for data_dict in (pv_mesh.cell_data, pv_mesh.point_data):
+                if stress_array_name not in data_dict:
+                    continue
+                source = np.asarray(data_dict[stress_array_name])
+                if source.ndim == 1 and source.size % 9 == 0:
+                    source = source.reshape(-1, 9)
+                if source.ndim != 2 or source.shape[1] != 9:
+                    continue
+
+                vm = self._von_mises_from_tensor(source)
+                data_dict[output_name] = vm
+                found_any = True
+                break
+
+        if not found_any:
+            raise ValueError(
+                f"No input mesh has a 9-component array named "
+                f"'{stress_array_name}' in point_data or cell_data."
+            )
+
+        # Invalidate cached MeshData so convert() picks up the new array.
+        self._cached_mesh_data = None
+        self.logger.info(
+            "Computed %s from %s on %d input mesh(es)",
+            output_name,
+            stress_array_name,
+            len(self.input_polydata),
+        )
+        return self
+
+    @staticmethod
+    def _von_mises_from_tensor(stress_tensor: np.ndarray) -> np.ndarray:
+        """Scalar von Mises stress from a row-major 9-component tensor field.
+
+        Args:
+            stress_tensor: Float array of shape ``(N, 9)``.
+
+        Returns:
+            Float32 array of shape ``(N,)``.
+        """
+        arr = np.asarray(stress_tensor, dtype=np.float64)
+        sxx, sxy, sxz = arr[:, 0], arr[:, 1], arr[:, 2]
+        syx, syy, syz = arr[:, 3], arr[:, 4], arr[:, 5]
+        szx, szy, szz = arr[:, 6], arr[:, 7], arr[:, 8]
+        sym_xy = 0.5 * (sxy + syx)
+        sym_yz = 0.5 * (syz + szy)
+        sym_zx = 0.5 * (sxz + szx)
+        deviatoric = 0.5 * ((sxx - syy) ** 2 + (syy - szz) ** 2 + (szz - sxx) ** 2)
+        shear = 3.0 * (sym_xy**2 + sym_yz**2 + sym_zx**2)
+        result: np.ndarray = np.sqrt(np.maximum(deviatoric + shear, 0.0)).astype(
+            np.float32
+        )
+        return result
+
     def convert(
         self,
         output_usd_file: str,
@@ -455,6 +570,14 @@ def convert(
             output_path.unlink()
             self.logger.debug(f"Removed existing file: {output_path}")
 
+        # USD caches layers globally by identifier, so a prior call in the
+        # same Python session can block CreateNew even after the file is
+        # gone. Evict any stale in-memory layer for this path.
+        stale_layer = Sdf.Layer.Find(str(output_path))
+        if stale_layer is not None:
+            stale_layer.Clear()
+            del stale_layer
+
         # Create USD stage
         stage = Usd.Stage.CreateNew(str(output_path))
         UsdGeom.SetStageMetersPerUnit(stage, self.settings.meters_per_unit)
@@ -489,6 +612,11 @@ def convert(
             # Single mesh (or time series) conversion
             self._convert_unified(stage, root_path, material_mgr, mesh_converter)
 
+        # Add a framing camera with tight near-clip so Omniverse Kit and other
+        # USD viewers can zoom close without geometry vanishing.
+        if add_framing_camera(stage) is None:
+            self.logger.debug("Skipped framing camera: no stage bounds available")
+
         # Save stage
         stage.Save()
         self.logger.info(f"Saved USD file: {output_path}")
@@ -619,6 +747,10 @@ def _convert_with_labels(
         for labeled_meshes in labeled_meshes_by_time:
             all_labels.update(labeled_meshes.keys())
 
+        # Track per-type Xforms/Scopes already defined so we only Define each
+        # once even when multiple labels share a type.
+        defined_type_groups: set[str] = set()
+
         # Convert each label separately
         for label_name in sorted(all_labels):
             self.logger.debug(f"Processing label: {label_name}")
@@ -638,11 +770,27 @@ def _convert_with_labels(
             if not label_mesh_sequence:
                 continue
 
-            # Create material for this label
-            material = self._create_material_from_colormap(f"{label_name}_material")
+            # Group labeled prims under an anatomy-type Xform so a 100-label
+            # output is structured (heart/, lung/, bone/, ...) instead of flat.
+            type_name = (
+                self.segmenter.label_to_type(label_name)
+                if self.segmenter is not None
+                else "Anatomy"
+            )
+            if type_name not in defined_type_groups:
+                UsdGeom.Xform.Define(stage, f"{root_path}/{type_name}")
+                UsdGeom.Scope.Define(stage, f"/World/Looks/{type_name}")
+                defined_type_groups.add(type_name)
+
+            # Create material for this label. The "/" in the material name
+            # makes MaterialManager author it at /World/Looks/{type}/{label}_material,
+            # so material paths mirror the mesh hierarchy.
+            material = self._create_material_from_colormap(
+                f"{type_name}/{label_name}_material"
+            )
 
             # Convert to USD
-            mesh_path = f"{root_path}/{label_name}"
+            mesh_path = f"{root_path}/{type_name}/{label_name}"
             if self._time_codes is not None:
                 label_time_codes = [self._time_codes[i] for i in label_frame_indices]
             else:
@@ -707,29 +855,35 @@ def _vtk_to_mesh_data(
         if self.color_by_array and self.color_by_array in vtk_mesh.point_data:
             colors = self._apply_colormap(vtk_mesh.point_data[self.color_by_array])
 
-        # Extract generic arrays
+        # Extract generic arrays from both point and cell data.
+        # Point arrays are vertex-interpolated; cell arrays are uniform
+        # (per-face) and are required for downstream colormap workflows that
+        # target simulation fields such as stress or strain.
         generic_arrays = []
-        for array_name in vtk_mesh.point_data.keys():
-            array_data = vtk_mesh.point_data[array_name]
-            num_components = int(array_data.shape[1] if array_data.ndim > 1 else 1)
-
-            # Determine data type
-            if array_data.dtype in [np.float32, np.float64]:
-                data_type = DataType.FLOAT
-            elif array_data.dtype in [np.int32, np.int64]:
-                data_type = DataType.INT
-            else:
-                data_type = DataType.FLOAT
-
-            generic_arrays.append(
-                GenericArray(
-                    name=array_name,
-                    data=array_data,
-                    num_components=num_components,
-                    data_type=data_type,
-                    interpolation="vertex",
+        for source, interpolation in (
+            (vtk_mesh.point_data, "vertex"),
+            (vtk_mesh.cell_data, "uniform"),
+        ):
+            for array_name in source.keys():
+                array_data = source[array_name]
+                num_components = int(array_data.shape[1] if array_data.ndim > 1 else 1)
+
+                if array_data.dtype in [np.float32, np.float64]:
+                    data_type = DataType.FLOAT
+                elif array_data.dtype in [np.int32, np.int64]:
+                    data_type = DataType.INT
+                else:
+                    data_type = DataType.FLOAT
+
+                generic_arrays.append(
+                    GenericArray(
+                        name=array_name,
+                        data=array_data,
+                        num_components=num_components,
+                        data_type=data_type,
+                        interpolation=interpolation,
+                    )
                 )
-            )
 
         return MeshData(
             points=points,
@@ -821,8 +975,11 @@ def _apply_colormap(self, scalar_data: np.ndarray) -> np.ndarray:
 
         colors_rgba = cmap(normalized)
 
-        # Return RGB (drop alpha)
-        return colors_rgba[:, :3].astype(np.float32)
+        # Return RGB (drop alpha). The intermediate variable pins the type so
+        # mypy doesn't lose track through .astype() — matplotlib's colormap
+        # return signature is loose.
+        colors_rgb: np.ndarray = colors_rgba[:, :3].astype(np.float32)
+        return colors_rgb
 
     def _create_material_from_colormap(self, name: str) -> MaterialData:
         """Create material based on colormap settings."""
diff --git a/src/physiomotion4d/data_download_tools.py b/src/physiomotion4d/data_download_tools.py
new file mode 100644
index 0000000..9af2a03
--- /dev/null
+++ b/src/physiomotion4d/data_download_tools.py
@@ -0,0 +1,128 @@
+"""
+Dataset download and verification helpers.
+
+Only Slicer-Heart-CT is downloaded automatically. Other datasets require
+manual download, and the verification helpers check the file layouts used by
+the repository tutorials, experiments, and tests.
+"""
+
+from __future__ import annotations
+
+import shutil
+import tempfile
+import urllib.request
+from pathlib import Path
+from typing import Union
+
+_DOWNLOAD_TIMEOUT_SECONDS = 60.0
+
+
+class DataDownloadTools:
+    """Download and verify optional PhysioMotion4D example datasets."""
+
+    SLICER_HEART_CT_URL = (
+        "https://github.com/SlicerHeart/SlicerHeart/releases/download/"
+        "TestingData/TruncalValve_4DCT.seq.nrrd"
+    )
+    SLICER_HEART_CT_FILENAME = "TruncalValve_4DCT.seq.nrrd"
+
+    @staticmethod
+    def DownloadSlicerHeartCTData(dirname: Union[str, Path]) -> Path:  # noqa: N802
+        """Download the Slicer-Heart-CT 4-D CT sample into ``dirname``.
+
+        Args:
+            dirname: Directory where ``TruncalValve_4DCT.seq.nrrd`` should live.
+
+        Returns:
+            Path to the downloaded or already-cached ``.seq.nrrd`` file.
+        """
+        data_dir = Path(dirname)
+        data_dir.mkdir(parents=True, exist_ok=True)
+
+        data_file = data_dir / DataDownloadTools.SLICER_HEART_CT_FILENAME
+        if data_file.exists() and data_file.stat().st_size > 0:
+            return data_file
+
+        # Stream to a unique temp file in the same directory with an explicit
+        # timeout, then atomically replace the target on success. The temp
+        # name is unique so concurrent callers do not clobber each other.
+        # Avoids partial files on interrupt and the indefinite hang that
+        # urlretrieve has without a timeout.
+        tmp_handle = tempfile.NamedTemporaryFile(
+            dir=str(data_dir),
+            prefix=f".{DataDownloadTools.SLICER_HEART_CT_FILENAME}.",
+            suffix=".tmp",
+            delete=False,
+        )
+        tmp_file = Path(tmp_handle.name)
+        try:
+            with (
+                urllib.request.urlopen(  # noqa: S310
+                    DataDownloadTools.SLICER_HEART_CT_URL,
+                    timeout=_DOWNLOAD_TIMEOUT_SECONDS,
+                ) as response,
+                tmp_handle as out,
+            ):
+                shutil.copyfileobj(response, out)
+            if tmp_file.stat().st_size == 0:
+                raise RuntimeError(
+                    f"Downloaded file is empty: {DataDownloadTools.SLICER_HEART_CT_URL}"
+                )
+            tmp_file.replace(data_file)
+        except BaseException:
+            tmp_handle.close()
+            if tmp_file.exists():
+                tmp_file.unlink()
+            raise
+        return data_file
+
+    @staticmethod
+    def VerifySlicerHeartCTData(dirname: Union[str, Path]) -> bool:  # noqa: N802
+        """Return True when Slicer-Heart-CT has the expected 4-D CT file."""
+        return (Path(dirname) / DataDownloadTools.SLICER_HEART_CT_FILENAME).is_file()
+
+    @staticmethod
+    def VerifyCHOPValve4DData(dirname: Union[str, Path]) -> bool:  # noqa: N802
+        """Return True when CHOP-Valve4D files referenced by the repo exist.
+
+        Accepted layouts are the CT volume used by Simpleware/model-to-patient
+        experiments and the valve time-series folders used by VTK-to-USD
+        experiments.
+        """
+        data_dir = Path(dirname)
+        has_ct_volume = any(
+            (data_dir / "CT" / filename).is_file()
+            for filename in ("RVOT28-Dias.nii.gz", "RVOT28-Dias.mha")
+        )
+        has_simpleware_parts = (data_dir / "CT" / "Simpleware" / "parts").is_dir()
+        has_alterra = (data_dir / "Alterra").is_dir() and any(
+            (data_dir / "Alterra").glob("*.vtk")
+        )
+        has_tpv25 = (data_dir / "TPV25").is_dir() and any(
+            (data_dir / "TPV25").glob("*.vtk")
+        )
+        return has_ct_volume or has_simpleware_parts or (has_alterra and has_tpv25)
+
+    @staticmethod
+    def VerifyDirLab4DCTData(dirname: Union[str, Path]) -> bool:  # noqa: N802
+        """Return True when a supported DirLab-4DCT case layout exists."""
+        data_dir = Path(dirname)
+        case1_dir = data_dir / "Case1"
+        has_case_dir_layout = case1_dir.is_dir() and any(case1_dir.glob("*.mha"))
+        has_case_dir_layout = has_case_dir_layout or (
+            case1_dir.is_dir() and any(case1_dir.glob("*.mhd"))
+        )
+
+        has_pack_layout = any(data_dir.glob("Case1Pack_T*.mhd")) or any(
+            data_dir.glob("Case1Pack_T*.mha")
+        )
+        return has_case_dir_layout or has_pack_layout
+
+    @staticmethod
+    def VerifyKCLHeartModelData(dirname: Union[str, Path]) -> bool:  # noqa: N802
+        """Return True when KCL-Heart-Model has its expected mesh inputs."""
+        data_dir = Path(dirname)
+        input_meshes_dir = data_dir / "input_meshes"
+        return (data_dir / "average_mesh.vtk").is_file() and any(
+            input_meshes_dir.glob("*.vtk")
+        )
diff --git a/src/physiomotion4d/notebook_utils.py b/src/physiomotion4d/notebook_utils.py
deleted file mode 100644
index fe09fc6..0000000
--- a/src/physiomotion4d/notebook_utils.py
+++ /dev/null
@@ -1,27 +0,0 @@
-"""
-Utilities for experiment notebooks (e.g. when run as tests by pytest).
-
-When experiment tests run via pytest with --run-experiments, the test runner sets
-PHYSIOMOTION_RUNNING_AS_TEST=1 so notebooks can use reduced parameters for fast runs.
-"""
-
-import os
-
-
-def running_as_test() -> bool:
-    """
-    True when the notebook is run as a test (e.g. by pytest experiment tests).
-
-    Use this to choose fast/small parameters (fewer iterations, fewer files, etc.)
-    so test runs complete in reasonable time. When False, use full parameters
-    for interactive or production runs.
-
-    Returns:
-        True if PHYSIOMOTION_RUNNING_AS_TEST is set to a truthy value
-        (1, true, yes, case-insensitive); False otherwise.
-    """
-    return os.environ.get("PHYSIOMOTION_RUNNING_AS_TEST", "").lower() in (
-        "1",
-        "true",
-        "yes",
-    )
diff --git a/src/physiomotion4d/register_images_ants.py b/src/physiomotion4d/register_images_ants.py
index c7ba4da..1142917 100644
--- a/src/physiomotion4d/register_images_ants.py
+++ b/src/physiomotion4d/register_images_ants.py
@@ -17,7 +17,6 @@
 import itk
 import numpy as np
 from numpy.typing import NDArray
-
 from physiomotion4d.register_images_base import RegisterImagesBase
 from physiomotion4d.transform_tools import TransformTools
 
@@ -640,7 +639,7 @@ def registration_method(
                 syn_metric=syn_metric,
                 use_histogram_matching=False,
                 mask_all_stages=True,
-                verbose=True,
+                verbose=False,
                 reg_iterations=self.number_of_iterations,
             )
         else:
@@ -652,7 +651,7 @@ def registration_method(
                 aff_metric=aff_metric,
                 syn_metric=syn_metric,
                 use_histogram_matching=False,
-                verbose=True,
+                verbose=False,
                 reg_iterations=self.number_of_iterations,
             )
 
diff --git a/src/physiomotion4d/segment_anatomy_base.py b/src/physiomotion4d/segment_anatomy_base.py
index cb2e652..ca2a788 100644
--- a/src/physiomotion4d/segment_anatomy_base.py
+++ b/src/physiomotion4d/segment_anatomy_base.py
@@ -12,6 +12,7 @@
 import numpy as np
 from itk import TubeTK as tube
 
+from physiomotion4d.anatomy_taxonomy import AnatomyTaxonomy
 from physiomotion4d.physiomotion4d_base import PhysioMotion4DBase
 
 
@@ -21,36 +22,43 @@ class SegmentAnatomyBase(PhysioMotion4DBase):
 
     This class implements preprocessing, postprocessing, and mask creation
     methods that are shared across different anatomy segmentation
-    implementations. It defines anatomical structure mappings and provides
-    utilities for image preprocessing, intensity rescaling, and mask generation.
-
-    The class maintains dictionaries of anatomical structure IDs for different
-    organ systems (heart, lungs, bones, vessels, etc.) and provides methods
-    to create binary masks for specific anatomical groups.
+    implementations. It owns an :class:`AnatomyTaxonomy` instance that
+    captures the group→organ structure (e.g. ``heart`` contains
+    ``atrial_appendage_left`` at id 61); subclasses populate it via
+    ``self.taxonomy.add_organ(...)`` and call
+    :meth:`_finalize_other_group` once they're done.
+
+    Extensibility
+    -------------
+    Each segmenter is free to define its own group names — the taxonomy does
+    not hard-code a fixed set. A new subclass adds groups by calling
+    ``self.taxonomy.add_organ(group_name, label_id, organ_name)`` for each
+    organ; the group is created lazily on first use. To assign a custom
+    OmniSurface look to a new group, register it in
+    :data:`physiomotion4d.usd_anatomy_tools.DEFAULT_RENDER_PARAMS` (see that
+    module's docstring). Groups without a registered look fall back to the
+    ``"other"`` entry, so they still render.
 
     Attributes:
-        target_spacing (float): Target isotropic spacing for resampling (default 1.5mm)
-        rescale_intensity_range (bool): Whether to rescale intensity values
-        contrast_threshold (int): Threshold for contrast agent detection (default 700)
-        all_mask_ids (dict): Dictionary mapping all anatomical structure IDs to names
-        heart_mask_ids (dict): Dictionary of cardiac structure IDs
-        lung_mask_ids (dict): Dictionary of pulmonary structure IDs
-        bone_mask_ids (dict): Dictionary of skeletal structure IDs
-        major_vessels_mask_ids (dict): Dictionary of major vessel IDs
-        contrast_mask_ids (dict): Dictionary of contrast-enhanced region IDs
-        soft_tissue_mask_ids (dict): Dictionary of soft tissue structure IDs
-        other_mask_ids (dict): Dictionary of remaining structure IDs
+        target_spacing (float): Target isotropic spacing for resampling.
+        rescale_intensity_range (bool): Whether to rescale intensity values.
+        contrast_threshold (int): Threshold for contrast agent detection.
+        taxonomy (AnatomyTaxonomy): Group→organ mapping shared with
+            :class:`physiomotion4d.USDAnatomyTools`.
     """
 
     def __init__(self, log_level: int | str = logging.INFO):
         """Initialize the SegmentAnatomyBase class.
 
-        Sets up default parameters for image preprocessing and anatomical
-        structure ID mappings. Subclasses should call this constructor and
-        then override the mask ID dictionaries with their specific mappings.
+        Sets up default parameters for image preprocessing and seeds the
+        anatomy taxonomy with the two base-class default organs (contrast
+        and soft_tissue). Subclasses should:
+
+        1. Add their organ groups via ``self.taxonomy.add_organ(...)``.
+        2. Call :meth:`_finalize_other_group` to fill in unclaimed ids.
 
         Args:
-            log_level: Logging level (default: logging.INFO)
+            log_level: Logging level (default: logging.INFO).
         """
         super().__init__(class_name=self.__class__.__name__, log_level=log_level)
 
@@ -63,50 +71,39 @@ def __init__(self, log_level: int | str = logging.INFO):
 
         self.contrast_threshold: int = 700
 
-        self.all_mask_ids: dict[int, str] = {}
-        self.heart_mask_ids: dict[int, str] = {}
-        self.major_vessels_mask_ids: dict[int, str] = {}
-        self.lung_mask_ids: dict[int, str] = {}
-        self.bone_mask_ids: dict[int, str] = {}
-        self.contrast_mask_ids: dict[int, str] = {135: "contrast"}
-        self.soft_tissue_mask_ids: dict[int, str] = {133: "soft_tissue"}
-        self.other_mask_ids: dict[int, str] = {}
+        # Single source of truth for the anatomy hierarchy. Subclasses
+        # populate this; USDAnatomyTools and ConvertVTKToUSD consume it.
+        self.taxonomy = AnatomyTaxonomy()
+        # Base-class default labels that downstream code relies on existing.
+        # Subclasses can override by adding the same id under a different
+        # group, or leave these in place.
+        self.taxonomy.add_organ("contrast", 135, "contrast")
+        self.taxonomy.add_organ("soft_tissue", 133, "soft_tissue")
+
+    def _finalize_other_group(self) -> None:
+        """Fill the ``other`` group with any unclaimed ids in [1, 256).
+
+        Subclasses call this at the end of ``__init__`` once they have
+        populated their specific groups. The consolidated all-labels view is
+        available via ``self.taxonomy.all_labels()``.
+        """
+        self.taxonomy.fill_other_group()
 
-        # Subclasses should call this function to complete the mask ID setup
-        # self.set_other_and_all_mask_ids()
+    def label_to_type(self, label_name: str) -> str:
+        """Return the anatomy group ('heart', 'lung', etc.) for a label name.
 
-    def set_other_and_all_mask_ids(self) -> None:
-        """Set the other mask IDs and consolidate all mask ID dictionaries.
+        Used by :class:`physiomotion4d.ConvertVTKToUSD` to group label-mode
+        mesh prims under per-type Xforms (e.g.
+        ``/World/{basename}/heart/{label_name}``). Delegates to the taxonomy.
 
-        Creates the 'other' category for any anatomical structures not classified
-        into the specific organ systems, then combines all individual mask ID
-        dictionaries into the master all_mask_ids dictionary.
+        Args:
+            label_name: Organ name (a value in the taxonomy's group organ dicts).
 
-        This method should be called after setting up the individual organ
-        system mask ID dictionaries in subclasses.
+        Returns:
+            The anatomy group name. Falls back to ``"other"`` for any label
+            the segmenter doesn't recognize.
         """
-        self.other_mask_ids = {id: f"other_{id}" for id in range(1, 256)}
-        for id in self.contrast_mask_ids.keys():
-            self.other_mask_ids.pop(id)
-        for id in self.soft_tissue_mask_ids.keys():
-            self.other_mask_ids.pop(id)
-        for id in self.heart_mask_ids.keys():
-            self.other_mask_ids.pop(id)
-        for id in self.major_vessels_mask_ids.keys():
-            self.other_mask_ids.pop(id)
-        for id in self.lung_mask_ids.keys():
-            self.other_mask_ids.pop(id)
-        for id in self.bone_mask_ids.keys():
-            self.other_mask_ids.pop(id)
-        self.all_mask_ids = {
-            **self.contrast_mask_ids,
-            **self.soft_tissue_mask_ids,
-            **self.heart_mask_ids,
-            **self.major_vessels_mask_ids,
-            **self.lung_mask_ids,
-            **self.bone_mask_ids,
-            **self.other_mask_ids,
-        }
+        return self.taxonomy.group_for_label(label_name)
 
     def set_target_spacing(self, target_spacing: float) -> None:
         """Set the target isotropic spacing for image resampling.
@@ -391,7 +388,7 @@ def segment_connected_component(
 
         label_arr = itk.GetArrayFromImage(labelmap_image)
         if labelmap_ids is None:
-            labelmap_ids = list(self.all_mask_ids.keys())
+            labelmap_ids = list(self.taxonomy.all_labels().keys())
         label_arr = np.isin(label_arr, labelmap_ids)
         label_image = itk.GetImageFromArray(label_arr.astype(np.int16))
         label_image.CopyInformation(labelmap_image)
@@ -469,18 +466,23 @@ def segment_contrast_agent(
             >>> contrast_labels = segmenter.segment_contrast_agent(preprocessed_image, base_labels)
         """
         thorasic_ids = (
-            list(self.heart_mask_ids.keys())
-            + list(self.lung_mask_ids.keys())
-            + list(self.major_vessels_mask_ids.keys())
+            list(self.taxonomy.labels_in_group("heart").keys())
+            + list(self.taxonomy.labels_in_group("lung").keys())
+            + list(self.taxonomy.labels_in_group("major_vessels").keys())
             + [0]
         )
+        contrast_ids = list(self.taxonomy.labels_in_group("contrast").keys())
+        if len(contrast_ids) == 0:
+            self.log_warning("No contrast-enhanced regions found in the labelmap")
+            return labelmap_image
+
         results_image = self.segment_connected_component(
             preprocessed_image,
             labelmap_image,
             lower_threshold=self.contrast_threshold,
             upper_threshold=4000,
             labelmap_ids=thorasic_ids,
-            mask_id=list(self.contrast_mask_ids.keys())[-1],
+            mask_id=contrast_ids[-1],
             use_mid_slice=True,
             hole_fill=3,
         )
@@ -501,77 +503,37 @@ def create_anatomy_group_masks(
             labelmap_image (itk.image): The detailed segmentation labelmap
 
         Returns:
-            dict[str, itk.image]: Dictionary of binary masks with keys:
-                - "lung": Pulmonary structures (lungs, trachea, airways)
-                - "heart": Cardiac structures (heart chambers, valves)
-                - "major_vessels": Large blood vessels (aorta, vena cava)
-                - "bone": Skeletal structures (ribs, vertebrae, sternum)
-                - "soft_tissue": Soft tissue organs (liver, kidneys, etc.)
-                - "other": Remaining anatomical structures
-                - "contrast": Contrast-enhanced regions
+            dict[str, itk.image]: Dictionary of binary masks keyed by group
+                name. Exactly one entry per group registered in
+                :attr:`taxonomy` (plus ``"other"``). The returned key set
+                is segmenter-specific — callers that need a particular
+                group should check membership (``"lung" in masks``) rather
+                than assume a fixed schema.
 
         Example:
             >>> masks = segmenter.create_anatomy_group_masks(labelmap)
-            >>> lung_mask = masks['lung']
-            >>> heart_mask = masks['heart']
+            >>> if "lung" in masks:
+            ...     lung_mask = masks["lung"]
         """
         labelmap_arr = itk.GetArrayFromImage(labelmap_image)
         other_mask_arr = np.where(labelmap_arr > 0, 1, 0)
 
-        lung_mask_arr = np.isin(labelmap_arr, list(self.lung_mask_ids.keys())).astype(
-            np.uint8
-        )
-        other_mask_arr = np.where(lung_mask_arr > 0, 0, other_mask_arr)
-
-        heart_mask_arr = np.isin(labelmap_arr, list(self.heart_mask_ids.keys())).astype(
-            np.uint8
-        )
-        other_mask_arr = np.where(heart_mask_arr > 0, 0, other_mask_arr)
-
-        major_vessels_mask_arr = np.isin(
-            labelmap_arr, list(self.major_vessels_mask_ids.keys())
-        ).astype(np.uint8)
-        other_mask_arr = np.where(major_vessels_mask_arr > 0, 0, other_mask_arr)
+        masks: dict[str, itk.image] = {}
+        for group_name in self.taxonomy.group_names():
+            if group_name == AnatomyTaxonomy.OTHER_GROUP:
+                continue
+            group_ids = list(self.taxonomy.labels_in_group(group_name).keys())
+            group_mask_arr = np.isin(labelmap_arr, group_ids).astype(np.uint8)
+            other_mask_arr = np.where(group_mask_arr > 0, 0, other_mask_arr)
+            group_mask = itk.GetImageFromArray(group_mask_arr)
+            group_mask.CopyInformation(labelmap_image)
+            masks[group_name] = group_mask
 
-        bone_mask_arr = np.isin(labelmap_arr, list(self.bone_mask_ids.keys())).astype(
-            np.uint8
-        )
-        other_mask_arr = np.where(bone_mask_arr > 0, 0, other_mask_arr)
-
-        soft_tissue_mask_arr = np.isin(
-            labelmap_arr, list(self.soft_tissue_mask_ids.keys())
-        ).astype(np.uint8)
-        other_mask_arr = np.where(soft_tissue_mask_arr > 0, 0, other_mask_arr)
-
-        contrast_mask_arr = np.isin(
-            labelmap_arr, list(self.contrast_mask_ids.keys())
-        ).astype(np.uint8)
-        other_mask_arr = np.where(contrast_mask_arr > 0, 0, other_mask_arr)
-
-        lung_mask = itk.GetImageFromArray(lung_mask_arr)
-        lung_mask.CopyInformation(labelmap_image)
-        heart_mask = itk.GetImageFromArray(heart_mask_arr)
-        heart_mask.CopyInformation(labelmap_image)
-        major_vessels_mask = itk.GetImageFromArray(major_vessels_mask_arr)
-        major_vessels_mask.CopyInformation(labelmap_image)
-        bone_mask = itk.GetImageFromArray(bone_mask_arr)
-        bone_mask.CopyInformation(labelmap_image)
-        soft_tissue_mask = itk.GetImageFromArray(soft_tissue_mask_arr)
-        soft_tissue_mask.CopyInformation(labelmap_image)
-        contrast_mask = itk.GetImageFromArray(contrast_mask_arr)
-        contrast_mask.CopyInformation(labelmap_image)
         other_mask = itk.GetImageFromArray(other_mask_arr.astype(np.uint8))
         other_mask.CopyInformation(labelmap_image)
+        masks[AnatomyTaxonomy.OTHER_GROUP] = other_mask
 
-        return {
-            "lung": lung_mask,
-            "heart": heart_mask,
-            "major_vessels": major_vessels_mask,
-            "bone": bone_mask,
-            "soft_tissue": soft_tissue_mask,
-            "other": other_mask,
-            "contrast": contrast_mask,
-        }
+        return masks
 
     def segmentation_method(self, preprocessed_image: itk.image) -> itk.image:
         """
@@ -668,13 +630,4 @@ def segment(
         )
         labelmap_image.CopyInformation(input_image)
 
-        return {
-            "labelmap": labelmap_image,
-            "lung": masks["lung"],
-            "heart": masks["heart"],
-            "major_vessels": masks["major_vessels"],
-            "bone": masks["bone"],
-            "soft_tissue": masks["soft_tissue"],
-            "other": masks["other"],
-            "contrast": masks["contrast"],
-        }
+        return {"labelmap": labelmap_image, **masks}
diff --git a/src/physiomotion4d/segment_chest_total_segmentator.py b/src/physiomotion4d/segment_chest_total_segmentator.py
index 14dc661..a51c207 100644
--- a/src/physiomotion4d/segment_chest_total_segmentator.py
+++ b/src/physiomotion4d/segment_chest_total_segmentator.py
@@ -32,20 +32,13 @@ class SegmentChestTotalSegmentator(SegmentAnatomyBase):
     combines the 'total' task (main organs and structures) with the 'body' task
     (body outline) to ensure complete coverage.
 
-    The class maintains specific ID mappings for:
-    - Heart structures (heart, atrial appendage, heart envelope)
-    - Major vessels (aorta, vena cava, carotid arteries, etc.)
-    - Lung structures (lung lobes, trachea, esophagus)
-    - Bone structures (vertebrae, ribs, sternum, skull)
-    - Soft tissue organs (liver, kidneys, spleen, etc.)
+    Anatomy groups (heart, lung, bone, major_vessels, soft_tissue) are
+    populated into :attr:`SegmentAnatomyBase.taxonomy` so downstream
+    consumers (``ConvertVTKToUSD``, ``USDAnatomyTools``) see a single,
+    consistent group→organ mapping.
 
     Attributes:
-        target_spacing (float): Target spacing set to 1.5mm for TotalSegmentator
-        heart_mask_ids (dict): Dictionary mapping heart structure IDs to names
-        major_vessels_mask_ids (dict): Dictionary mapping vessel IDs to names
-        lung_mask_ids (dict): Dictionary mapping lung structure IDs to names
-        bone_mask_ids (dict): Dictionary mapping bone structure IDs to names
-        soft_tissue_mask_ids (dict): Dictionary mapping soft tissue IDs to names
+        target_spacing (float): Target spacing set to 1.5mm for TotalSegmentator.
 
     Example:
         >>> segmenter = SegmentChestTotalSegmentator()
@@ -57,9 +50,10 @@ class SegmentChestTotalSegmentator(SegmentAnatomyBase):
     def __init__(self, log_level: int | str = logging.INFO):
         """Initialize the TotalSegmentator-based chest segmentation.
 
-        Sets up the TotalSegmentator-specific anatomical structure ID mappings
-        and processing parameters. The target spacing is set to 1.5mm which
-        provides a good balance between accuracy and processing speed.
+        Populates :attr:`SegmentAnatomyBase.taxonomy` with the
+        TotalSegmentator class index space, then calls
+        :meth:`SegmentAnatomyBase._finalize_other_group` so unclaimed ids end
+        up in the ``other`` group.
 
         Args:
             log_level: Logging level (default: logging.INFO)
@@ -68,133 +62,149 @@ def __init__(self, log_level: int | str = logging.INFO):
 
         self.target_spacing = 1.5
 
-        self.heart_mask_ids = {
-            51: "heart",
-            61: "atrial_appendage_left",
-            140: "heart_envelop",
-        }
-
-        self.major_vessels_mask_ids = {
-            52: "aorta",
-            53: "pulmonary_vein",
-            54: "brachiocephalic_trunk",
-            55: "right_subclavian_artery",
-            56: "left_subclavian_artery",
-            57: "common_carotid_artery_right",
-            58: "common_carotid_artery_left",
-            59: "brachiocephalic_vein_left",
-            60: "brachiocephalic_vein_right",
-            62: "superior_vena_cava",
-            63: "inferior_vena_cava",
-            120: "lung_vessels",
-        }
-
-        self.lung_mask_ids = {
-            10: "lung_upper_lobe_left",
-            11: "lung_lower_lobe_left",
-            12: "lung_upper_lobe_right",
-            13: "lung_middle_lobe_right",
-            14: "lung_lower_lobe_right",
-        }
-
-        self.bone_mask_ids = {
-            26: "vertebra_S1",
-            27: "vertebra_L5",
-            28: "vertebra_L4",
-            29: "vertebrae_L3",
-            30: "vertebrae_L2",
-            31: "vertebrae_L1",
-            32: "vertebrae_T12",
-            33: "vertebrae_T11",
-            34: "vertebrae_T10",
-            35: "vertebrae_T9",
-            36: "vertebrae_T8",
-            37: "vertebrae_T7",
-            38: "vertebrae_T6",
-            39: "vertebrae_T5",
-            40: "vertebrae_T4",
-            41: "vertebrae_T3",
-            42: "vertebrae_T2",
-            43: "vertebrae_T1",
-            44: "vertebrae_C7",
-            45: "vertebrae_C6",
-            46: "vertebrae_C5",
-            47: "vertebrae_C4",
-            48: "vertebrae_C3",
-            49: "vertebrae_C2",
-            50: "vertebrae_C1",
-            69: "humerus_left",
-            70: "humerus_right",
-            71: "scapula_left",
-            72: "scapula_right",
-            73: "clavicula_left",
-            74: "clavicula_right",
-            75: "femur_left",
-            76: "femur_right",
-            77: "hip_left",
-            78: "hip_right",
-            91: "skull",
-            92: "rib_left_1",
-            93: "rib_left_2",
-            94: "rib_left_3",
-            95: "rib_left_4",
-            96: "rib_left_5",
-            97: "rib_left_6",
-            98: "rib_left_7",
-            99: "rib_left_8",
-            100: "rib_left_9",
-            101: "rib_left_10",
-            102: "rib_left_11",
-            103: "rib_left_12",
-            104: "rib_right_1",
-            105: "rib_right_2",
-            106: "rib_right_3",
-            107: "rib_right_4",
-            108: "rib_right_5",
-            109: "rib_right_6",
-            110: "rib_right_7",
-            111: "rib_right_8",
-            112: "rib_right_9",
-            113: "rib_right_10",
-            114: "rib_right_11",
-            115: "rib_right_12",
-            116: "sternum",
-            117: "costal_cartilages",
-        }
-
-        self.soft_tissue_mask_ids = {
-            1: "spleen",
-            2: "kidney_right",
-            3: "kidney_left",
-            4: "gallbladder",
-            5: "liver",
-            6: "stomach",
-            7: "pancreas",
-            8: "adrenal_gland_right",
-            9: "adrenal_gland_left",
-            17: "thyroid_gland",
-            18: "small_bowel",
-            19: "duodenum",
-            20: "colon",
-            21: "urinary_bladder",
-            22: "prostate",
-            25: "sacrum",
-            80: "gluteus_maximus_left",
-            81: "gluteus_maximus_right",
-            82: "gluteus_medius_left",
-            83: "gluteus_medius_right",
-            84: "gluteus_minimus_left",
-            85: "gluteus_minimus_right",
-            90: "brain",
-            15: "esophagus",
-            16: "trachea",
-            133: "soft_tissue",
-        }
-
-        # From Base Class
-        # self.contrast_mask_ids = { 135: "contrast" }
-
-        self.set_other_and_all_mask_ids()
+        # TotalSegmentator class indices, grouped by anatomy. Contrast (135)
+        # and the generic soft_tissue label (133) come from the base class.
+        for group_name, organs in (
+            (
+                "heart",
+                {
+                    51: "heart",
+                    61: "atrial_appendage_left",
+                    140: "heart_envelop",
+                },
+            ),
+            (
+                "major_vessels",
+                {
+                    52: "aorta",
+                    53: "pulmonary_vein",
+                    54: "brachiocephalic_trunk",
+                    55: "right_subclavian_artery",
+                    56: "left_subclavian_artery",
+                    57: "common_carotid_artery_right",
+                    58: "common_carotid_artery_left",
+                    59: "brachiocephalic_vein_left",
+                    60: "brachiocephalic_vein_right",
+                    62: "superior_vena_cava",
+                    63: "inferior_vena_cava",
+                },
+            ),
+            (
+                "lung",
+                {
+                    10: "lung_upper_lobe_left",
+                    11: "lung_lower_lobe_left",
+                    12: "lung_upper_lobe_right",
+                    13: "lung_middle_lobe_right",
+                    14: "lung_lower_lobe_right",
+                    120: "lung_arteries",
+                    121: "lung_veins",
+                    122: "lung_airways",
+                    123: "lung_airways_wall",
+                },
+            ),
+            (
+                "bone",
+                {
+                    26: "vertebra_S1",
+                    27: "vertebra_L5",
+                    28: "vertebra_L4",
+                    29: "vertebrae_L3",
+                    30: "vertebrae_L2",
+                    31: "vertebrae_L1",
+                    32: "vertebrae_T12",
+                    33: "vertebrae_T11",
+                    34: "vertebrae_T10",
+                    35: "vertebrae_T9",
+                    36: "vertebrae_T8",
+                    37: "vertebrae_T7",
+                    38: "vertebrae_T6",
+                    39: "vertebrae_T5",
+                    40: "vertebrae_T4",
+                    41: "vertebrae_T3",
+                    42: "vertebrae_T2",
+                    43: "vertebrae_T1",
+                    44: "vertebrae_C7",
+                    45: "vertebrae_C6",
+                    46: "vertebrae_C5",
+                    47: "vertebrae_C4",
+                    48: "vertebrae_C3",
+                    49: "vertebrae_C2",
+                    50: "vertebrae_C1",
+                    69: "humerus_left",
+                    70: "humerus_right",
+                    71: "scapula_left",
+                    72: "scapula_right",
+                    73: "clavicula_left",
+                    74: "clavicula_right",
+                    75: "femur_left",
+                    76: "femur_right",
+                    77: "hip_left",
+                    78: "hip_right",
+                    91: "skull",
+                    92: "rib_left_1",
+                    93: "rib_left_2",
+                    94: "rib_left_3",
+                    95: "rib_left_4",
+                    96: "rib_left_5",
+                    97: "rib_left_6",
+                    98: "rib_left_7",
+                    99: "rib_left_8",
+                    100: "rib_left_9",
+                    101: "rib_left_10",
+                    102: "rib_left_11",
+                    103: "rib_left_12",
+                    104: "rib_right_1",
+                    105: "rib_right_2",
+                    106: "rib_right_3",
+                    107: "rib_right_4",
+                    108: "rib_right_5",
+                    109: "rib_right_6",
+                    110: "rib_right_7",
+                    111: "rib_right_8",
+                    112: "rib_right_9",
+                    113: "rib_right_10",
+                    114: "rib_right_11",
+                    115: "rib_right_12",
+                    116: "sternum",
+                    117: "costal_cartilages",
+                },
+            ),
+            (
+                "soft_tissue",
+                {
+                    1: "spleen",
+                    2: "kidney_right",
+                    3: "kidney_left",
+                    4: "gallbladder",
+                    5: "liver",
+                    6: "stomach",
+                    7: "pancreas",
+                    8: "adrenal_gland_right",
+                    9: "adrenal_gland_left",
+                    17: "thyroid_gland",
+                    18: "small_bowel",
+                    19: "duodenum",
+                    20: "colon",
+                    21: "urinary_bladder",
+                    22: "prostate",
+                    25: "sacrum",
+                    80: "gluteus_maximus_left",
+                    81: "gluteus_maximus_right",
+                    82: "gluteus_medius_left",
+                    83: "gluteus_medius_right",
+                    84: "gluteus_minimus_left",
+                    85: "gluteus_minimus_right",
+                    90: "brain",
+                    15: "esophagus",
+                    16: "trachea",
+                },
+            ),
+        ):
+            for label_id, organ_name in organs.items():
+                self.taxonomy.add_organ(group_name, label_id, organ_name)
+
+        self._finalize_other_group()
 
     def segmentation_method(self, preprocessed_image: itk.image) -> itk.image:
         """
@@ -252,22 +262,25 @@ def segmentation_method(self, preprocessed_image: itk.image) -> itk.image:
             mask1 = labelmap_arr1 == 0
             mask2 = labelmap_arr2 > 0
             mask = mask1 & mask2
-            final_arr = np.where(
-                mask, list(self.soft_tissue_mask_ids.keys())[-1], labelmap_arr1
+            # The base class registers (133, "soft_tissue") as a generic
+            # placeholder; use that id to fill the body mask where
+            # TotalSegmentator's 'total' task didn't classify anything.
+            soft_tissue_id = next(
+                label_id
+                for label_id, name in self.taxonomy.labels_in_group(
+                    "soft_tissue"
+                ).items()
+                if name == "soft_tissue"
             )
+            final_arr = np.where(mask, soft_tissue_id, labelmap_arr1)
 
-            mask3 = labelmap_arr3 == 1
-            mask3_img = itk.image_from_array(mask3.astype(np.uint8))
-            mask3_img.CopyInformation(preprocessed_image)
-            imMath = itk.TubeTK.ImageMath.New(mask3_img)
-            imMath.Dilate(1, 1, 0)
-            imMath.Erode(1, 1, 0)
-            mask3_img = imMath.GetOutputUChar()
-            mask3 = itk.array_from_image(mask3_img)
-            final_arr = np.where(mask3, 120, final_arr)  # lung vessels
-            mask3 = labelmap_arr3 == 2
-            final_arr = np.where(mask3, 16, final_arr)  # trachea
-
+            # labelmap_arr3 contains: 1=arteries, 2=veins, 3=airways, 4=airways_wall
+            final_arr = np.where(labelmap_arr3 == 1, 120, final_arr)  # lung arteries
+            final_arr = np.where(labelmap_arr3 == 2, 121, final_arr)  # lung veins
+            final_arr = np.where(labelmap_arr3 == 3, 122, final_arr)  # lung airways
+            final_arr = np.where(
+                labelmap_arr3 == 4, 123, final_arr
+            )  # lung airways wall
             # To create an ITK image, we save the result and read it back with
             # ITK. This correctly handles the coordinate system and data
             # layout conversions.
diff --git a/src/physiomotion4d/segment_heart_simpleware.py b/src/physiomotion4d/segment_heart_simpleware.py
index 7ba6ddc..2b39a18 100644
--- a/src/physiomotion4d/segment_heart_simpleware.py
+++ b/src/physiomotion4d/segment_heart_simpleware.py
@@ -40,11 +40,13 @@ class SegmentHeartSimpleware(SegmentAnatomyBase):
     - Major vessels (aorta, pulmonary artery)
 
     Attributes:
-        target_spacing (float): Target spacing set to 1.0mm for Simpleware
-        heart_mask_ids (dict): Dictionary mapping heart structure IDs to names
-        major_vessels_mask_ids (dict): Dictionary mapping vessel IDs to names
-        simpleware_exe_path (str): Path to Simpleware Medical executable
-        simpleware_script_path (str): Path to Simpleware Python script
+        target_spacing (float): Target spacing set to 1.0mm for Simpleware.
+        simpleware_exe_path (str): Path to Simpleware Medical executable.
+        simpleware_script_path (str): Path to Simpleware Python script.
+
+    The heart and major-vessel labels populated by this class are accessed
+    through the inherited :attr:`SegmentAnatomyBase.taxonomy`
+    (``taxonomy.labels_in_group("heart")`` etc.).
 
     Example:
         >>> segmenter = SegmentHeartSimpleware()
@@ -69,41 +71,39 @@ def __init__(self, log_level: int | str = logging.INFO):
 
         self.target_spacing = 1.0
 
-        # Heart structure IDs for Simpleware Medical ASCardio output
-        # These IDs should match the output from the ASCardio module
-        self.heart_mask_ids = {
-            1: "left_ventricle",
-            2: "right_ventricle",
-            3: "left_atrium",
-            4: "right_atrium",
-            5: "myocardium",
-            6: "heart",
-        }
-
-        # Major vessel IDs for Simpleware Medical ASCardio output
-        self.major_vessels_mask_ids = {
-            7: "aorta",
-            8: "pulmonary_artery",
-            9: "right_coronary_artery",
-            10: "left_coronary_artery",
-        }
-
-        # Lung structures are not segmented by ASCardio
-        self.lung_mask_ids = {}
-
-        # Bone structures are not segmented by ASCardio
-        self.bone_mask_ids = {}
-
-        # Soft tissue structures are not segmented by ASCardio
-        # (will be filled in by base class 'other' category)
-        self.soft_tissue_mask_ids = {}
-
-        # From Base Class
-        # self.contrast_mask_ids = {135: "contrast"}
+        # Heart and major-vessel labels from Simpleware Medical ASCardio.
+        # Lung / bone / soft_tissue are not segmented by ASCardio; they will
+        # be folded into the 'other' group by _finalize_other_group().
+        # Contrast (135) and soft_tissue (133) defaults are inherited from
+        # SegmentAnatomyBase.
+        for group_name, organs in (
+            (
+                "heart",
+                {
+                    1: "left_ventricle",
+                    2: "right_ventricle",
+                    3: "left_atrium",
+                    4: "right_atrium",
+                    5: "myocardium",
+                    6: "heart",
+                },
+            ),
+            (
+                "major_vessels",
+                {
+                    7: "aorta",
+                    8: "pulmonary_artery",
+                    9: "right_coronary_artery",
+                    10: "left_coronary_artery",
+                },
+            ),
+        ):
+            for label_id, organ_name in organs.items():
+                self.taxonomy.add_organ(group_name, label_id, organ_name)
 
         self._trim_mask = False
 
-        self.set_other_and_all_mask_ids()
+        self._finalize_other_group()
 
         # Path to Simpleware Medical console executable
         self.simpleware_exe_path = "C:/Program Files/Synopsys/Simpleware Medical/X-2025.06/ConsoleSimplewareMedical.exe"
@@ -268,7 +268,7 @@ def segmentation_method(self, preprocessed_image: itk.image) -> itk.image:
             sz = sz[::-1]
             labelmap_array = np.zeros(sz, dtype=np.uint8)
             interior_array = np.zeros(sz, dtype=np.uint8)
-            for mask_id, mask_name in self.all_mask_ids.items():
+            for mask_id, mask_name in self.taxonomy.all_labels().items():
                 output_file = os.path.join(tmp_dir, f"mask_{mask_name}.mhd")
                 if os.path.exists(output_file):
                     mask_image = itk.imread(output_file)
@@ -283,17 +283,29 @@ def segmentation_method(self, preprocessed_image: itk.image) -> itk.image:
                         labelmap_array == 0, mask_array, labelmap_array
                     )
 
+            # landmarks.csv is optional: Simpleware Medical's ASCardio module
+            # writes it for some valid inputs and omits it for others (e.g.
+            # very small ROIs or unsupported acquisitions). Treat its
+            # absence as "no landmarks for this case" rather than a hard
+            # failure, so callers that only need the labelmap still succeed.
             landmarks_file = os.path.join(tmp_dir, "landmarks.csv")
             self.landmarks.clear()
-            with open(landmarks_file, newline="", encoding="utf-8-sig") as fh:
-                next(fh)  # skip line 1 (file header)
-                for row in csv.DictReader(fh):
-                    coords = row["Measurement"].replace(" mm", "").split(",")
-                    self.landmarks[row["Name"]] = (
-                        float(coords[0]),
-                        float(coords[1]),
-                        float(coords[2]),
-                    )
+            if os.path.exists(landmarks_file):
+                with open(landmarks_file, newline="", encoding="utf-8-sig") as fh:
+                    next(fh)  # skip line 1 (file header)
+                    for row in csv.DictReader(fh):
+                        coords = row["Measurement"].replace(" mm", "").split(",")
+                        self.landmarks[row["Name"]] = (
+                            float(coords[0]),
+                            float(coords[1]),
+                            float(coords[2]),
+                        )
+            else:
+                self.log_warning(
+                    "Simpleware did not write landmarks.csv (looked at %s); "
+                    "continuing with an empty landmark set.",
+                    landmarks_file,
+                )
 
             interior_image = itk.GetImageFromArray(interior_array.astype(np.uint8))
             interior_image.CopyInformation(preprocessed_image)
diff --git a/src/physiomotion4d/simpleware_medical/README.md b/src/physiomotion4d/simpleware_medical/README.md
index 6c20b41..dbbc238 100644
--- a/src/physiomotion4d/simpleware_medical/README.md
+++ b/src/physiomotion4d/simpleware_medical/README.md
@@ -18,7 +18,7 @@ The integration enables PhysioMotion4D to leverage Simpleware Medical's ASCardio
 
 ## Current Status
 
-**✅ FUNCTIONAL**: This integration works as expected.
+**FUNCTIONAL**: This integration works as expected.
 
 ### How It Works
 
diff --git a/src/physiomotion4d/test_tools.py b/src/physiomotion4d/test_tools.py
index 2961704..2240eb4 100644
--- a/src/physiomotion4d/test_tools.py
+++ b/src/physiomotion4d/test_tools.py
@@ -9,9 +9,10 @@
 from __future__ import annotations
 
 import logging
+import os
 import shutil
 from pathlib import Path
-from typing import Any, Literal, Optional
+from typing import Any, Literal, Optional, cast
 
 import itk
 import numpy as np
@@ -43,35 +44,41 @@ class TestTools(PhysioMotion4DBase):
 
     def __init__(
         self,
-        results_dir: Path,
-        baselines_dir: Path,
         class_name: str,
+        results_dir: Optional[Path] = None,
+        baselines_dir: Optional[Path] = None,
         *,
-        results_output_dir: Optional[Path] = None,
         log_level: int = logging.INFO,
     ) -> None:
         """Initialize test helpers.
 
         Args:
-            results_dir: Root directory for result artifacts.
-            baselines_dir: Root directory for baseline artifacts.
-            class_name: Subdirectory name for baselines and, by default, results.
-            results_output_dir: Optional exact directory for result artifacts.
-                When set, results are read and written there instead of
-                ``results_dir / class_name`` while baselines still use
-                ``baselines_dir / class_name``.
+            class_name: Identifier used for the default results/baselines
+                subdirectory and the logger name. Callers that supply
+                ``results_dir`` or ``baselines_dir`` explicitly are
+                responsible for including ``class_name`` in the path if they
+                want a per-test subdirectory.
+            results_dir: Exact directory for written result artifacts. Used
+                as-is. Defaults to ``<repo>/tests/results/<class_name>`` when
+                ``None``.
+            baselines_dir: Exact directory for baseline artifacts. Used
+                as-is. Defaults to ``<repo>/tests/baselines/<class_name>``
+                when ``None``.
             log_level: Logging level.
         """
         super().__init__(class_name=class_name, log_level=log_level)
 
-        self._results_dir = (
-            results_output_dir
-            if results_output_dir is not None
-            else results_dir / class_name
-        )
+        self._tests_dir = _REPO_ROOT / "tests"
+        if results_dir is not None:
+            self._results_dir = results_dir
+        else:
+            self._results_dir = self._tests_dir / "results" / class_name
         self._results_dir.mkdir(parents=True, exist_ok=True)
 
-        self._baselines_dir = baselines_dir / class_name
+        if baselines_dir is not None:
+            self._baselines_dir = baselines_dir
+        else:
+            self._baselines_dir = self._tests_dir / "baselines" / class_name
         self._baselines_dir.mkdir(parents=True, exist_ok=True)
 
         self._last_image_per_pixel_absolute_error_tol: float | None = None
@@ -92,6 +99,25 @@ def __init__(
             None  # itk.Transform (type depends on template)
         )
 
+    @staticmethod
+    def running_as_test() -> bool:
+        """
+        True when the script is run as a test (e.g. by pytest experiment tests).
+
+        Use this to choose fast/small parameters (fewer iterations, fewer files, etc.)
+        so test runs complete in reasonable time. When False, use full parameters
+        for interactive or production runs.
+
+        Returns:
+            True if PHYSIOMOTION_RUNNING_AS_TEST is set to a truthy value
+            (1, true, yes, case-insensitive); False otherwise.
+        """
+        return os.environ.get("PHYSIOMOTION_RUNNING_AS_TEST", "").lower() in (
+            "1",
+            "true",
+            "yes",
+        )
+
     def image_pass_fail_and_pixels_above_tolerance(self) -> tuple[bool, int]:
         """
         Return (pass, value) for number of pixels above tolerance from the most
@@ -409,6 +435,75 @@ def save_screenshot_mesh(
         self.log_info("Screenshot saved: %s", output_path)
         return output_path
 
+    def save_screenshot_openusd(
+        self,
+        usd_file: str | Path,
+        filename: str,
+        *,
+        prim_path: str = "/World",
+        time_code: Optional[float] = None,
+    ) -> Path:
+        """Render USD mesh geometry off-screen and save a PNG.
+
+        The scene is loaded through :meth:`USDTools.load_usd_as_vtk` into a
+        PyVista mesh, rendered with a fixed isometric camera and fixed
+        ``800 x 600`` window, and centered automatically by PyVista.
+
+        Args:
+            usd_file: USD file to render.
+            filename: Output PNG filename, relative to the result artifact dir.
+            prim_path: USD prim path to render. Defaults to ``/World``.
+            time_code: Optional animation time code. ``None`` renders default
+                values and falls back to the first authored mesh point sample.
+
+        Returns:
+            Absolute path to the saved PNG.
+        """
+        import os
+        import sys
+
+        import pyvista as pv
+
+        from physiomotion4d.usd_tools import USDTools
+
+        # On headless Linux runners VTK needs an X server or off-screen GL
+        # context. If DISPLAY is already provided (e.g. xvfb-run wrapping
+        # pytest), trust it. Otherwise try pv.start_xvfb() and let failures
+        # surface — silently swallowing them previously caused VTK to
+        # segfault inside Plotter.screenshot() on GitHub Actions.
+        xvfb_started = False
+        if sys.platform.startswith("linux") and not os.environ.get("DISPLAY"):
+            pv.start_xvfb()
+            xvfb_started = True
+
+        try:
+            output_path = self._results_dir / filename
+            mesh = USDTools().load_usd_as_vtk(
+                usd_file, prim_path=prim_path, time_code=time_code
+            )
+            plotter = pv.Plotter(off_screen=True, window_size=[800, 600])
+            try:
+                if "openusd_rgb" in mesh.point_data:
+                    plotter.add_mesh(mesh, scalars="openusd_rgb", rgb=True)
+                else:
+                    plotter.add_mesh(mesh, color="red")
+                plotter.camera_position = "iso"
+                # pyvista wraps reset_camera in a descriptor that mypy can't
+                # resolve as a bound method (see pyvista issue with _Wrapped).
+                # Cast through Any to keep the call expression valid for mypy
+                # without resorting to a # type: ignore comment, which flips
+                # between "missing self" and "unused ignore" depending on
+                # check scope.
+                cast(Any, plotter).reset_camera()
+                plotter.screenshot(str(output_path))
+            finally:
+                plotter.close()
+        finally:
+            if xvfb_started and hasattr(pv, "stop_xvfb"):
+                pv.stop_xvfb()
+        self.log_info("OpenUSD screenshot saved: %s", output_path)
+        return output_path
+
     def save_screenshot_image_slice(
         self,
         image: Any,  # itk.Image, axes X Y Z in RAS world space
diff --git a/src/physiomotion4d/transform_tools.py b/src/physiomotion4d/transform_tools.py
index 9b18afd..2a22aa2 100644
--- a/src/physiomotion4d/transform_tools.py
+++ b/src/physiomotion4d/transform_tools.py
@@ -13,6 +13,7 @@
 
 import logging
 from collections.abc import Sequence
+from pathlib import Path
 from typing import TypeAlias, cast
 
 try:
@@ -28,10 +29,11 @@
 import SimpleITK as sitk
 import vtk
 from numpy.typing import NDArray
-from pxr import Gf, Usd, UsdGeom
+from pxr import Gf, Sdf, Usd, UsdGeom
 
 from physiomotion4d.image_tools import ImageTools
 from physiomotion4d.physiomotion4d_base import PhysioMotion4DBase
+from physiomotion4d.vtk_to_usd import add_framing_camera
 
 FloatArray: TypeAlias = NDArray[np.float32] | NDArray[np.float64]
 
@@ -895,6 +897,17 @@ def convert_itk_transform_to_usd_visualization(
             size[2] // subsample_factor,
         ]
 
+        # Remove any existing file and evict any stale in-memory USD layer.
+        # USD caches layers globally by identifier, so a prior call in the
+        # same Python session can block CreateNew even after the file is gone.
+        output_path = Path(output_filename)
+        if output_path.exists():
+            output_path.unlink()
+        stale_layer = Sdf.Layer.Find(str(output_path))
+        if stale_layer is not None:
+            stale_layer.Clear()
+            del stale_layer
+
         # Create USD stage
         stage = Usd.Stage.CreateNew(output_filename)
 
@@ -930,6 +943,9 @@ def convert_itk_transform_to_usd_visualization(
                 "Must be 'arrows' or 'flowlines'."
             )
 
+        # Framing camera with tight near-clip for Omniverse Kit viewer ergonomics.
+        add_framing_camera(stage)
+
         # Save the stage
         stage.Save()
         self.log_info("Created USD visualization: %s", output_filename)
diff --git a/src/physiomotion4d/usd_anatomy_tools.py b/src/physiomotion4d/usd_anatomy_tools.py
index 829d326..e703eba 100644
--- a/src/physiomotion4d/usd_anatomy_tools.py
+++ b/src/physiomotion4d/usd_anatomy_tools.py
@@ -1,220 +1,258 @@
 """
 This module contains the USDAnatomyTools class, which is used to enhance
 the anatomy meshes in a USD file.
+
+Extensibility
+-------------
+The default OmniSurface look for each anatomy group/organ lives in the
+module-level :data:`DEFAULT_RENDER_PARAMS` dict. A new segmenter that
+introduces a new group (e.g. ``"brain"``, ``"tumor"``) can register a
+matching look in one of three ways:
+
+1. **Globally**, before instantiating any ``USDAnatomyTools``::
+
+       from physiomotion4d.usd_anatomy_tools import DEFAULT_RENDER_PARAMS
+       DEFAULT_RENDER_PARAMS["brain"] = {"name": "Brain", ...}
+
+   Every subsequent ``USDAnatomyTools`` instance picks up the new entry.
+
+2. **Per-instance**, after construction::
+
+       tools = USDAnatomyTools(stage)
+       tools.render_params["brain"] = {"name": "Brain", ...}
+
+3. **By subclassing**, overriding ``__init__`` to populate
+   ``self.render_params`` with project-specific defaults.
+
+Group lookup falls back to ``render_params["other"]`` when a group has no
+registered entry, so any group present in the segmenter's
+:class:`physiomotion4d.AnatomyTaxonomy` will still render *something*.
 """
 
 import logging
 from typing import Any, Mapping
 
-from pxr import Sdf, Usd, UsdGeom, UsdShade
+from pxr import Sdf, UsdGeom, UsdShade
 
 from physiomotion4d.physiomotion4d_base import PhysioMotion4DBase
 
+# Default OmniSurface render parameters keyed by group name (matching
+# :class:`physiomotion4d.AnatomyTaxonomy.group_names`) and by organ-level
+# overrides (e.g. ``liver``, ``spleen``, ``kidney_left``). ``enhance_meshes``
+# consults an organ-level entry first, then falls back to the containing
+# group's entry, and finally to ``"other"``. Module-level so CLIs and tests
+# can enumerate the supported types without constructing a USD stage.
+DEFAULT_RENDER_PARAMS: dict[str, dict[str, Any]] = {
+    "heart": {
+        "name": "Heart",
+        "enable_diffuse_transmission": True,
+        "diffuse_reflection_weight": 0.25,
+        "diffuse_reflection_color": (0.2, 0.01, 0.01),
+        "diffuse_reflection_roughness": 0.5,
+        "metalness": 0.0,
+        "specular_reflection_weight": 0.5,
+        "specular_reflection_roughness": 0.5,
+        "subsurface_transmission_color": (0.8, 0.4, 0.4),
+        "subsurface_scattering_color": (0.8, 0.4, 0.4),
+        "subsurface_weight": 0.1,
+        "subsurface_scale": 2.0,
+        "coat_weight": 0.0,
+    },
+    "lung": {
+        "name": "Lung",
+        "enable_diffuse_transmission": True,
+        "diffuse_reflection_weight": 0.125,
+        "diffuse_reflection_color": (0.34, 0.0, 0.0),
+        "diffuse_reflection_roughness": 0.6,
+        "metalness": 0.0,
+        "specular_reflection_weight": 0.5,
+        "specular_reflection_roughness": 0.5,
+        "subsurface_transmission_color": (0.9, 0.7, 0.7),
+        "subsurface_scattering_color": (0.9, 0.7, 0.7),
+        "subsurface_weight": 0.1,
+        "subsurface_scale": 0.2,
+        "coat_weight": 0.0,
+    },
+    "bone": {
+        "name": "Bone",
+        "enable_diffuse_transmission": True,
+        "diffuse_reflection_weight": 0.2,
+        "diffuse_reflection_color": (0.8, 0.8, 0.9),
+        "diffuse_reflection_roughness": 0.3,
+        "metalness": 0.0,
+        "specular_reflection_weight": 0.5,
+        "specular_reflection_roughness": 0.5,
+        "subsurface_transmission_color": (0.95, 0.9, 0.7),
+        "subsurface_scattering_color": (0.95, 0.9, 0.7),
+        "subsurface_weight": 0.03,
+        "subsurface_scale": 0.05,
+        "coat_weight": 0.0,
+    },
+    "major_vessels": {
+        "name": "Major_Vessels",
+        "enable_diffuse_transmission": True,
+        "diffuse_reflection_weight": 0.35,
+        "diffuse_reflection_color": (0.2, 0.01, 0.01),
+        "diffuse_reflection_roughness": 0.3,
+        "metalness": 0.0,
+        "specular_reflection_weight": 0.5,
+        "specular_reflection_roughness": 0.5,
+        "subsurface_transmission_color": (0.7, 0.15, 0.18),
+        "subsurface_scattering_color": (0.7, 0.15, 0.18),
+        "subsurface_weight": 0.09,
+        "subsurface_scale": 0.22,
+        "coat_weight": 0.12,
+    },
+    "contrast": {
+        "name": "Contrast",
+        "enable_diffuse_transmission": True,
+        "diffuse_reflection_weight": 0.25,
+        "diffuse_reflection_color": (0.2, 0.01, 0.01),
+        "diffuse_reflection_roughness": 0.5,
+        "metalness": 0.0,
+        "specular_reflection_weight": 0.5,
+        "specular_reflection_roughness": 0.5,
+        "subsurface_transmission_color": (0.9, 0.4, 0.4),
+        "subsurface_scattering_color": (0.9, 0.4, 0.4),
+        "subsurface_weight": 0.1,
+        "subsurface_scale": 2.0,
+        "coat_weight": 0.0,
+    },
+    "soft_tissue": {
+        "name": "Soft_Tissue",
+        "enable_diffuse_transmission": True,
+        "diffuse_reflection_weight": 0.17,
+        "diffuse_reflection_color": (0.7, 0.5, 0.4),
+        "diffuse_reflection_roughness": 0.5,
+        "metalness": 0.0,
+        "specular_reflection_weight": 0.5,
+        "specular_reflection_roughness": 0.5,
+        "subsurface_transmission_color": (0.95, 0.9, 0.9),
+        "subsurface_scattering_color": (0.95, 0.9, 0.9),
+        "subsurface_weight": 0.08,
+        "subsurface_scale": 0.3,
+        "coat_weight": 0.12,
+    },
+    "other": {
+        "name": "Other",
+        "enable_diffuse_transmission": True,
+        "diffuse_reflection_weight": 0.1,
+        "diffuse_reflection_color": (0.7, 0.5, 0.4),
+        "diffuse_reflection_roughness": 0.5,
+        "metalness": 0.0,
+        "specular_reflection_weight": 0.5,
+        "specular_reflection_roughness": 0.5,
+        "subsurface_transmission_color": (0.95, 0.5, 0.4),
+        "subsurface_scattering_color": (0.95, 0.5, 0.4),
+        "subsurface_weight": 0.1,
+        "subsurface_scale": 0.3,
+        "coat_weight": 0.12,
+    },
+    # Organ-level overrides. enhance_meshes consults these by organ
+    # name *before* falling back to the containing group's params, so
+    # liver/spleen/kidney get their dedicated look despite being in
+    # the soft_tissue group of the taxonomy.
+    "liver": {
+        "name": "Liver",
+        "enable_diffuse_transmission": True,
+        "diffuse_reflection_weight": 0.1,
+        "diffuse_reflection_color": (0.16, 0.01, 0.01),
+        "diffuse_reflection_roughness": 0.4,
+        "metalness": 0.0,
+        "specular_reflection_weight": 0.01,
+        "specular_reflection_roughness": 0.5,
+        "subsurface_transmission_color": (0.7, 0.2, 0.15),
+        "subsurface_scattering_color": (0.7, 0.2, 0.15),
+        "subsurface_weight": 0.08,
+        "subsurface_scale": 0.15,
+        "coat_weight": 0.01,
+    },
+    "spleen": {
+        "name": "Spleen",
+        "enable_diffuse_transmission": True,
+        "diffuse_reflection_weight": 0.1,
+        "diffuse_reflection_color": (0.45, 0.08, 0.15),
+        "diffuse_reflection_roughness": 0.4,
+        "metalness": 0.0,
+        "specular_reflection_weight": 0.5,
+        "specular_reflection_roughness": 0.5,
+        "subsurface_transmission_color": (0.6, 0.15, 0.22),
+        "subsurface_scattering_color": (0.6, 0.15, 0.22),
+        "subsurface_weight": 0.08,
+        "subsurface_scale": 0.15,
+        "coat_weight": 0.1,
+    },
+    "kidney_right": {
+        "name": "Kidney",
+        "enable_diffuse_transmission": True,
+        "diffuse_reflection_weight": 0.1,
+        "diffuse_reflection_color": (0.45, 0.13, 0.12),
+        "diffuse_reflection_roughness": 0.35,
+        "metalness": 0.0,
+        "specular_reflection_weight": 0.5,
+        "specular_reflection_roughness": 0.5,
+        "subsurface_transmission_color": (0.6, 0.18, 0.15),
+        "subsurface_scattering_color": (0.6, 0.18, 0.15),
+        "subsurface_weight": 0.085,
+        "subsurface_scale": 0.18,
+        "coat_weight": 0.1,
+    },
+}
+# kidney_left and kidney share the same look as kidney_right ("kidney" is a
+# convenience alias kept for the CLI and other generic callers).
+DEFAULT_RENDER_PARAMS["kidney_left"] = DEFAULT_RENDER_PARAMS["kidney_right"]
+DEFAULT_RENDER_PARAMS["kidney"] = DEFAULT_RENDER_PARAMS["kidney_right"]
+
 
 class USDAnatomyTools(PhysioMotion4DBase):
-    """
-    This class is used to enhance the appearance of anatomy meshes in a USD
-    file.
+    """Apply OmniSurface materials to anatomy mesh prims in a USD stage.
+
+    The instance attribute :attr:`render_params` is initialized from the
+    module-level :data:`DEFAULT_RENDER_PARAMS` (deep copy per instance, so
+    in-place edits stay local). See the module docstring for how to add new
+    groups/organs.
     """
 
     def __init__(self, stage: Any, log_level: int | str = logging.INFO) -> None:
         """Initialize USDAnatomyTools.
 
         Args:
-            stage: USD stage to work with
-            log_level: Logging level (default: logging.INFO)
+            stage: USD stage to work with. May be ``None`` when the instance
+                is only used for color look-ups (e.g. via
+                :meth:`get_anatomy_diffuse_color`).
+            log_level: Logging level (default: logging.INFO).
         """
         super().__init__(class_name=self.__class__.__name__, log_level=log_level)
         self.stage = stage
-
-        self.heart_params = {
-            "name": "Heart",
-            "enable_diffuse_transmission": True,
-            "diffuse_reflection_weight": 0.25,
-            "diffuse_reflection_color": (0.2, 0.01, 0.01),
-            "diffuse_reflection_roughness": 0.5,
-            "metalness": 0.0,
-            "specular_reflection_weight": 0.5,
-            "specular_reflection_roughness": 0.5,
-            "subsurface_transmission_color": (0.8, 0.4, 0.4),
-            "subsurface_scattering_color": (0.8, 0.4, 0.4),
-            "subsurface_weight": 0.1,
-            "subsurface_scale": 2.0,
-            "coat_weight": 0.0,
-        }
-        self.lung_params = {
-            "name": "Lung",
-            "enable_diffuse_transmission": True,
-            "diffuse_reflection_weight": 0.125,
-            "diffuse_reflection_color": (0.34, 0.0, 0.0),
-            "diffuse_reflection_roughness": 0.6,
-            "metalness": 0.0,
-            "specular_reflection_weight": 0.5,
-            "specular_reflection_roughness": 0.5,
-            "subsurface_transmission_color": (0.9, 0.7, 0.7),
-            "subsurface_scattering_color": (0.9, 0.7, 0.7),
-            "subsurface_weight": 0.1,
-            "subsurface_scale": 0.2,
-            "coat_weight": 0.0,
-        }
-        self.bone_params = {
-            "name": "Bone",
-            "enable_diffuse_transmission": True,
-            "diffuse_reflection_weight": 0.2,
-            "diffuse_reflection_color": (0.8, 0.8, 0.9),
-            "diffuse_reflection_roughness": 0.3,
-            "metalness": 0.0,
-            "specular_reflection_weight": 0.5,
-            "specular_reflection_roughness": 0.5,
-            "subsurface_transmission_color": (0.95, 0.9, 0.7),
-            "subsurface_scattering_color": (0.95, 0.9, 0.7),
-            "subsurface_weight": 0.03,
-            "subsurface_scale": 0.05,
-            "coat_weight": 0.0,
-        }
-        self.major_vessels_params = {
-            "name": "Major_Vessels",
-            "enable_diffuse_transmission": True,
-            "diffuse_reflection_weight": 0.35,
-            "diffuse_reflection_color": (0.2, 0.01, 0.01),
-            "diffuse_reflection_roughness": 0.3,
-            "metalness": 0.0,
-            "specular_reflection_weight": 0.5,
-            "specular_reflection_roughness": 0.5,
-            "subsurface_transmission_color": (0.7, 0.15, 0.18),
-            "subsurface_scattering_color": (0.7, 0.15, 0.18),
-            "subsurface_weight": 0.09,
-            "subsurface_scale": 0.22,
-            "coat_weight": 0.12,
-        }
-        self.contrast_params = {
-            "name": "Contrast",
-            "enable_diffuse_transmission": True,
-            "diffuse_reflection_weight": 0.25,
-            "diffuse_reflection_color": (0.2, 0.01, 0.01),
-            "diffuse_reflection_roughness": 0.5,
-            "metalness": 0.0,
-            "specular_reflection_weight": 0.5,
-            "specular_reflection_roughness": 0.5,
-            "subsurface_transmission_color": (0.9, 0.4, 0.4),
-            "subsurface_scattering_color": (0.9, 0.4, 0.4),
-            "subsurface_weight": 0.1,
-            "subsurface_scale": 2.0,
-            "coat_weight": 0.0,
-        }
-        self.soft_tissue_params = {
-            "name": "Soft_Tissue",
-            "enable_diffuse_transmission": True,
-            "diffuse_reflection_weight": 0.17,
-            "diffuse_reflection_color": (0.7, 0.5, 0.4),
-            "diffuse_reflection_roughness": 0.5,
-            "metalness": 0.0,
-            "specular_reflection_weight": 0.5,
-            "specular_reflection_roughness": 0.5,
-            "subsurface_transmission_color": (0.95, 0.9, 0.9),
-            "subsurface_scattering_color": (0.95, 0.9, 0.9),
-            "subsurface_weight": 0.08,
-            "subsurface_scale": 0.3,
-            "coat_weight": 0.12,
-        }
-        self.other_params = {
-            "name": "Other",
-            "enable_diffuse_transmission": True,
-            "diffuse_reflection_weight": 0.1,
-            "diffuse_reflection_color": (0.7, 0.5, 0.4),
-            "diffuse_reflection_roughness": 0.5,
-            "metalness": 0.0,
-            "specular_reflection_weight": 0.5,
-            "specular_reflection_roughness": 0.5,
-            "subsurface_transmission_color": (0.95, 0.5, 0.4),
-            "subsurface_scattering_color": (0.95, 0.5, 0.4),
-            "subsurface_weight": 0.1,
-            "subsurface_scale": 0.3,
-            "coat_weight": 0.12,
-        }
-        self.liver_params = {
-            "name": "Liver",
-            "enable_diffuse_transmission": True,
-            "diffuse_reflection_weight": 0.1,
-            "diffuse_reflection_color": (0.16, 0.01, 0.01),
-            "diffuse_reflection_roughness": 0.4,
-            "metalness": 0.0,
-            "specular_reflection_weight": 0.01,
-            "specular_reflection_roughness": 0.5,
-            "subsurface_transmission_color": (0.7, 0.2, 0.15),
-            "subsurface_scattering_color": (0.7, 0.2, 0.15),
-            "subsurface_weight": 0.08,
-            "subsurface_scale": 0.15,
-            "coat_weight": 0.01,
-        }
-        self.spleen_params = {
-            "name": "Spleen",
-            "enable_diffuse_transmission": True,
-            "diffuse_reflection_weight": 0.1,
-            "diffuse_reflection_color": (0.45, 0.08, 0.15),
-            "diffuse_reflection_roughness": 0.4,
-            "metalness": 0.0,
-            "specular_reflection_weight": 0.5,
-            "specular_reflection_roughness": 0.5,
-            "subsurface_transmission_color": (0.6, 0.15, 0.22),
-            "subsurface_scattering_color": (0.6, 0.15, 0.22),
-            "subsurface_weight": 0.08,
-            "subsurface_scale": 0.15,
-            "coat_weight": 0.1,
-        }
-        self.kidney_params = {
-            "name": "Kidney",
-            "enable_diffuse_transmission": True,
-            "diffuse_reflection_weight": 0.1,
-            "diffuse_reflection_color": (0.45, 0.13, 0.12),
-            "diffuse_reflection_roughness": 0.35,
-            "metalness": 0.0,
-            "specular_reflection_weight": 0.5,
-            "specular_reflection_roughness": 0.5,
-            "subsurface_transmission_color": (0.6, 0.18, 0.15),
-            "subsurface_scattering_color": (0.6, 0.18, 0.15),
-            "subsurface_weight": 0.085,
-            "subsurface_scale": 0.18,
-            "coat_weight": 0.1,
-        }
-
-        # Map anatomy type name (CLI/workflow) to params for apply_anatomy_material_to_mesh
-        self._anatomy_params_by_type: Mapping[str, Mapping[str, Any]] = {
-            "heart": self.heart_params,
-            "lung": self.lung_params,
-            "bone": self.bone_params,
-            "major_vessels": self.major_vessels_params,
-            "contrast": self.contrast_params,
-            "soft_tissue": self.soft_tissue_params,
-            "other": self.other_params,
-            "liver": self.liver_params,
-            "spleen": self.spleen_params,
-            "kidney": self.kidney_params,
+        # Per-instance copy so per-instance mutations don't leak into other
+        # USDAnatomyTools instances.
+        self.render_params: dict[str, dict[str, Any]] = {
+            key: dict(params) for key, params in DEFAULT_RENDER_PARAMS.items()
         }
 
     def get_anatomy_types(self) -> list[str]:
-        """Return list of supported anatomy type names for apply_anatomy_material_to_mesh."""
-        return list(self._anatomy_params_by_type.keys())
+        """Return list of registered render-param keys (groups + organ overrides)."""
+        return list(self.render_params.keys())
 
     def get_anatomy_diffuse_color(
         self, anatomy_type: str
     ) -> tuple[float, float, float]:
-        """Return the diffuse reflection RGB color for the given anatomy type.
+        """Return the diffuse reflection RGB color for the given group/organ.
 
         This accessor does not require a USD stage and may be called on an instance
         created with ``stage=None`` purely for color look-up purposes.
 
         Args:
-            anatomy_type: One of: heart, lung, bone, major_vessels, contrast,
-                soft_tissue, other, liver, spleen, kidney.
+            anatomy_type: A registered render-params key (e.g. ``"heart"``,
+                ``"lung"``, ``"liver"``).
 
         Returns:
             RGB tuple of floats in ``[0, 1]``.
 
         Raises:
-            ValueError: If *anatomy_type* is not supported.
+            ValueError: If *anatomy_type* is not registered.
         """
-        params = self._anatomy_params_by_type.get(anatomy_type.lower())
+        params = self.render_params.get(anatomy_type.lower())
         if params is None:
             raise ValueError(
                 f"Unknown anatomy_type '{anatomy_type}'. "
@@ -228,13 +266,13 @@ def apply_anatomy_material_to_mesh(self, mesh_path: str, anatomy_type: str) -> N
 
         Args:
             mesh_path: USD path to the mesh prim (e.g. "/World/Meshes/MyMesh").
-            anatomy_type: One of: heart, lung, bone, major_vessels, contrast,
-                soft_tissue, other, liver, spleen, kidney.
+            anatomy_type: A registered render-params key (group or organ
+                override). See :meth:`get_anatomy_types`.
 
         Raises:
-            ValueError: If mesh_path is invalid or anatomy_type is not supported.
+            ValueError: If mesh_path is invalid or anatomy_type is not registered.
         """
-        params = self._anatomy_params_by_type.get(anatomy_type.lower())
+        params = self.render_params.get(anatomy_type.lower())
         if params is None:
             raise ValueError(
                 f"Unknown anatomy_type '{anatomy_type}'. "
@@ -319,117 +357,46 @@ def apply_anatomy_material_to_prim(
         binding_api.Bind(material)
 
     def enhance_meshes(self, segmentator: Any) -> None:
-        """Find and enhance all heart meshes"""
-
-        heart_mask_ids = list(segmentator.heart_mask_ids.values())
-        major_vessels_mask_ids = list(segmentator.major_vessels_mask_ids.values())
-        lung_mask_ids = list(segmentator.lung_mask_ids.values())
-        bone_mask_ids = list(segmentator.bone_mask_ids.values())
-        contrast_mask_ids = list(segmentator.contrast_mask_ids.values())
-        soft_tissue_mask_ids = list(segmentator.soft_tissue_mask_ids.values())
-        other_mask_ids = list(segmentator.other_mask_ids.values())
-
-        liver_mask_ids = ["liver"]
-        spleen_mask_ids = ["spleen"]
-        kidney_mask_ids = ["kidney_right", "kidney_left"]
-
-        # Safely remove items if they exist
-        for item in ["liver", "spleen", "kidney_right", "kidney_left"]:
-            if item in soft_tissue_mask_ids:
-                soft_tissue_mask_ids.remove(item)
-
-        list_of_mask_ids = [
-            heart_mask_ids,
-            major_vessels_mask_ids,
-            lung_mask_ids,
-            bone_mask_ids,
-            contrast_mask_ids,
-            soft_tissue_mask_ids,
-            other_mask_ids,
-            liver_mask_ids,
-            spleen_mask_ids,
-            kidney_mask_ids,
-        ]
-        list_of_params = [
-            self.heart_params,
-            self.major_vessels_params,
-            self.lung_params,
-            self.bone_params,
-            self.contrast_params,
-            self.soft_tissue_params,
-            self.other_params,
-            self.liver_params,
-            self.spleen_params,
-            self.kidney_params,
-        ]
-
-        editor = Usd.NamespaceEditor(self.stage)
-        prims = []
-        for prim in self.stage.Traverse():
-            prims.append(prim)
-        for prim in prims:
-            prim_name = str(prim.GetPrimPath())
-            prim_name = prim_name.split("/")[-1]
-            prim_sub_name = "_".join(prim_name.split("_")[1:])
+        """Apply per-organ OmniSurface materials to every matching mesh prim.
 
-            anatomy_params = None
-            anatomy_prim_found = False
-            for mask_ids, params in zip(list_of_mask_ids, list_of_params):
-                if prim_name in mask_ids or prim_sub_name in mask_ids:
-                    anatomy_params = params
-                    anatomy_prim_found = True
-                    break
-
-            if anatomy_prim_found:
-                assert anatomy_params is not None
-                mesh_prim = UsdGeom.Mesh(prim)
-                transform_prim = UsdGeom.Xform(prim)
-                if transform_prim and not mesh_prim:
-                    current_prim_path = str(prim.GetPrimPath())
-                    root_prim_path = "/".join(current_prim_path.split("/")[:-1])
-                    self.log_debug("Root prim path: %s", root_prim_path)
-                    anatomy_prim_path = "/".join([root_prim_path, "Anatomy"])
-                    self.log_debug("   Anatomy prim path: %s", anatomy_prim_path)
-                    if not self.stage.GetPrimAtPath(anatomy_prim_path):
-                        UsdGeom.Xform.Define(self.stage, anatomy_prim_path)
-                    anatomy_prim_path = "/".join(
-                        [root_prim_path, "Anatomy", f"Group_{anatomy_params['name']}"]
-                    )
-                    if not self.stage.GetPrimAtPath(anatomy_prim_path):
-                        UsdGeom.Xform.Define(self.stage, anatomy_prim_path)
-                    remaining_prim_path = prim.GetName()
-                    anatomy_prim_path = "/".join(
-                        [
-                            root_prim_path,
-                            "Anatomy",
-                            f"Group_{anatomy_params['name']}",
-                            remaining_prim_path,
-                        ]
-                    )
-                    anatomy_prim_path = Sdf.Path(anatomy_prim_path)
-                    self.log_debug("   Current prim path: %s", current_prim_path)
-                    self.log_debug("   Anatomy prim path: %s", anatomy_prim_path)
-                    editor.MovePrimAtPath(
-                        current_prim_path,
-                        anatomy_prim_path,
-                    )
-                    editor.ApplyEdits()
+        Walks the segmenter's :class:`AnatomyTaxonomy` and applies a material
+        to each mesh prim whose leaf name matches an organ name in any group.
+        An organ-level entry in :attr:`render_params` (e.g. ``"liver"``,
+        ``"spleen"``, ``"kidney_left"``) takes precedence over the entry for
+        the containing group.
+
+        Anatomy grouping is performed upstream by ConvertVTKToUSD, which
+        writes labeled prims under ``/World/{basename}/{type}/{label_name}``.
+        This method only needs to apply materials; it does not move prims.
+
+        Args:
+            segmentator: A :class:`SegmentAnatomyBase` instance whose
+                ``taxonomy`` attribute holds the group/organ structure.
+        """
+        taxonomy = segmentator.taxonomy
+
+        # Build organ_name -> render params dict in one pass. Organ-level
+        # overrides win over the containing group's params; if neither is
+        # registered, fall back to the "other" entry (always present in
+        # DEFAULT_RENDER_PARAMS, so the lookup is safe).
+        organ_params: dict[str, dict[str, Any]] = {}
+        default_params: dict[str, Any] = self.render_params["other"]
+        for group_name in taxonomy.group_names():
+            group_params = self.render_params.get(group_name, default_params)
+            for organ_name in taxonomy.labels_in_group(group_name).values():
+                organ_params[organ_name] = self.render_params.get(
+                    organ_name, group_params
+                )
 
         for prim in self.stage.Traverse():
-            prim_name = str(prim.GetPrimPath())
-            prim_name = prim_name.split("/")[-1]
+            mesh_prim = UsdGeom.Mesh(prim)
+            if not mesh_prim:
+                continue
+            prim_name = prim.GetName()
+            # ConvertVTKToUSD may prefix prim names with an index like
+            # "frame0_<organ>"; accept the suffix as a match too.
             prim_sub_name = "_".join(prim_name.split("_")[1:])
-
-            anatomy_params = None
-            anatomy_prim_found = False
-            for mask_ids, params in zip(list_of_mask_ids, list_of_params):
-                if prim_name in mask_ids or prim_sub_name in mask_ids:
-                    anatomy_params = params
-                    anatomy_prim_found = True
-                    break
-
-            if anatomy_prim_found:
-                assert anatomy_params is not None
-                mesh_prim = UsdGeom.Mesh(prim)
-                if mesh_prim:
-                    self.apply_anatomy_material_to_prim(prim, anatomy_params)
+            params = organ_params.get(prim_name) or organ_params.get(prim_sub_name)
+            if params is None:
+                continue
+            self.apply_anatomy_material_to_prim(prim, params)
diff --git a/src/physiomotion4d/usd_tools.py b/src/physiomotion4d/usd_tools.py
index 76123c4..05b9173 100644
--- a/src/physiomotion4d/usd_tools.py
+++ b/src/physiomotion4d/usd_tools.py
@@ -14,12 +14,15 @@
 
 import logging
 from collections.abc import Sequence
-from typing import Any
+from pathlib import Path
+from typing import Any, cast
 
 import numpy as np
-from pxr import Gf, Usd, UsdGeom, UsdShade
+import pyvista as pvtk
+from pxr import Gf, Sdf, Usd, UsdGeom, UsdShade
 
 from physiomotion4d.physiomotion4d_base import PhysioMotion4DBase
+from physiomotion4d.vtk_to_usd import add_framing_camera
 
 
 class USDTools(PhysioMotion4DBase):
@@ -66,6 +69,152 @@ def __init__(self, log_level: int | str = logging.INFO) -> None:
         """
         super().__init__(class_name=self.__class__.__name__, log_level=log_level)
 
+    def load_usd_as_vtk(
+        self,
+        usd_file: str | Path,
+        prim_path: str = "/World",
+        time_code: float | None = None,
+    ) -> pvtk.PolyData:
+        """Load USD mesh geometry as a PyVista ``PolyData``.
+
+        Evaluates mesh points at ``time_code``, applies each mesh prim's
+        local-to-world transform, and stores RGB colors in
+        ``point_data['openusd_rgb']``. Authored ``displayColor`` is used when
+        available; otherwise points are colored red. Coordinates are returned
+        in the USD stage coordinate system.
+
+        Args:
+            usd_file: Path to a USD file.
+            prim_path: Root prim path to traverse. Defaults to ``/World``.
+            time_code: Optional time code for animated meshes. ``None`` reads
+                default values and falls back to the first authored point time
+                sample.
+
+        Returns:
+            A merged PyVista ``PolyData`` containing all mesh prims under
+            ``prim_path``.
+
+        Raises:
+            FileNotFoundError: If ``usd_file`` does not exist.
+            ValueError: If the stage, prim path, or mesh geometry is invalid.
+        """
+        usd_path = Path(usd_file)
+        if not usd_path.exists():
+            raise FileNotFoundError(f"USD file not found: {usd_path}")
+
+        stage = Usd.Stage.Open(str(usd_path))
+        if stage is None:
+            raise ValueError(f"Could not open USD file: {usd_path}")
+
+        root_prim = stage.GetPrimAtPath(prim_path)
+        if not root_prim.IsValid():
+            raise ValueError(f"USD prim path not found: {prim_path}")
+
+        if time_code is None:
+            usd_time = Usd.TimeCode.Default()
+        else:
+            usd_time = Usd.TimeCode(time_code)
+        xform_cache = UsdGeom.XformCache(usd_time)
+
+        meshes: list[pvtk.PolyData] = []
+        for prim in Usd.PrimRange(root_prim):
+            if not prim.IsA(UsdGeom.Mesh):
+                continue
+
+            mesh = UsdGeom.Mesh(prim)
+            points_value = mesh.GetPointsAttr().Get(usd_time)
+            if points_value is None and time_code is None:
+                point_samples = mesh.GetPointsAttr().GetTimeSamples()
+                if point_samples:
+                    sample_time = Usd.TimeCode(point_samples[0])
+                    points_value = mesh.GetPointsAttr().Get(sample_time)
+            if points_value is None or len(points_value) == 0:
+                continue
+
+            face_counts = mesh.GetFaceVertexCountsAttr().Get(usd_time)
+            face_indices = mesh.GetFaceVertexIndicesAttr().Get(usd_time)
+            if face_counts is None or face_indices is None:
+                continue
+
+            world_matrix = xform_cache.GetLocalToWorldTransform(prim)
+            # Vectorize the local-to-world transform: USD's Gf.Matrix4d uses
+            # the convention `world_point = local_point_row_vec * M`, where
+            # the matrix is row-major and the translation row is the last
+            # row. Building a (N, 4) homogeneous-point block and multiplying
+            # once is dramatically faster than calling Transform() per point
+            # for large meshes.
+            mat_array = np.array(
+                [[float(world_matrix[i][j]) for j in range(4)] for i in range(4)],
+                dtype=np.float64,
+            )
+            local_points = np.asarray(points_value, dtype=np.float64)
+            homogeneous = np.empty((local_points.shape[0], 4), dtype=np.float64)
+            homogeneous[:, :3] = local_points
+            homogeneous[:, 3] = 1.0
+            points = (homogeneous @ mat_array)[:, :3].astype(np.float32)
+            if len(points) == 0:
+                continue
+
+            faces: list[int] = []
+            index_offset = 0
+            for count in face_counts:
+                count_int = int(count)
+                faces.append(count_int)
+                faces.extend(
+                    int(face_indices[index_offset + i]) for i in range(count_int)
+                )
+                index_offset += count_int
+            if not faces:
+                continue
+
+            pv_mesh = pvtk.PolyData(points, np.asarray(faces, dtype=np.int64))
+            rgb = self._usd_display_color(mesh, len(points), usd_time)
+            pv_mesh.point_data["openusd_rgb"] = rgb
+            meshes.append(pv_mesh)
+
+        if not meshes:
+            raise ValueError(f"No mesh geometry found in {usd_path} under {prim_path}")
+
+        if len(meshes) == 1:
+            return meshes[0]
+        # pvtk.merge is loosely typed (it can return several DataSet subclasses
+        # depending on inputs). All callers here pass PolyData with
+        # merge_points=False, so the result is always PolyData.
+        return cast(pvtk.PolyData, pvtk.merge(meshes, merge_points=False))
+
+    @staticmethod
+    def _usd_display_color(
+        mesh: UsdGeom.Mesh,
+        n_points: int,
+        time_code: Usd.TimeCode,
+    ) -> np.ndarray:
+        """Return point RGB colors from ``displayColor`` or the red fallback."""
+        fallback = np.tile(np.array([[255, 0, 0]], dtype=np.uint8), (n_points, 1))
+        primvar = UsdGeom.PrimvarsAPI(mesh).GetPrimvar("displayColor")
+        if not primvar:
+            return fallback
+
+        color_value = primvar.Get(time_code)
+        if color_value is None:
+            color_value = primvar.Get()
+        if color_value is None or len(color_value) == 0:
+            return fallback
+
+        colors = np.asarray(color_value, dtype=np.float32)
+        if colors.ndim != 2 or colors.shape[1] < 3:
+            return fallback
+
+        colors = colors[:, :3]
+        interpolation = primvar.GetInterpolation()
+        if interpolation in (UsdGeom.Tokens.constant, "constant"):
+            colors = np.tile(colors[0], (n_points, 1))
+        elif len(colors) == 1:
+            colors = np.tile(colors[0], (n_points, 1))
+        elif len(colors) != n_points:
+            return fallback
+
+        return np.asarray(np.clip(colors, 0.0, 1.0) * 255.0, dtype=np.uint8)
+
     def get_subtree_bounding_box(
         self, prim: UsdGeom.Xform
     ) -> tuple[Gf.Vec3f, Gf.Vec3f]:
@@ -246,6 +395,9 @@ def save_usd_file_arrangement(
             #                     prim.GetPrimPath(),
             #                 )
 
+        # Framing camera with tight near-clip for Omniverse Kit viewer ergonomics.
+        add_framing_camera(new_stage)
+
         self.log_info("Exporting stage...")
         new_stage.Export(new_stage_name)
 
@@ -285,6 +437,17 @@ def merge_usd_files(
             ...     'complete_anatomy.usd', ['heart_dynamic.usd', 'lungs_static.usd', 'skeleton.usd']
             ... )
         """
+        # Remove any existing file and evict any stale in-memory USD layer.
+        # USD caches layers globally by identifier, so a prior call in the
+        # same Python session can block CreateNew even after the file is gone.
+        output_path = Path(output_filename)
+        if output_path.exists():
+            output_path.unlink()
+        stale_layer = Sdf.Layer.Find(str(output_path))
+        if stale_layer is not None:
+            stale_layer.Clear()
+            del stale_layer
+
         # Create new stage with meters as units (standard USD configuration)
         stage = Usd.Stage.CreateNew(output_filename)
         stage.SetMetadata("metersPerUnit", 1.0)
@@ -436,6 +599,9 @@ def _copy_prim(src_prim: Any, target_path: str) -> None:
                 frames_per_second,
             )
 
+        # Framing camera with tight near-clip for Omniverse Kit viewer ergonomics.
+        add_framing_camera(stage)
+
         # Save with USDA format
         # stage.GetRootLayer().Export(output_path, args=['--usdFormat', 'usda'])
         stage.Export(output_filename)
@@ -561,6 +727,9 @@ def merge_usd_files_flattened(
                 output_stage.SetFramesPerSecond(frames_per_second)
                 self.log_info("Set output FramesPerSecond: %s", frames_per_second)
 
+        # Framing camera with tight near-clip for Omniverse Kit viewer ergonomics.
+        add_framing_camera(output_stage)
+
         # Export the flattened layer with corrected metadata
         self.log_info("Exporting to %s", output_filename)
         output_stage.Export(output_filename)
@@ -1266,11 +1435,8 @@ def _uniform_to_vertex_scalar(
         Returns:
             np.ndarray: Scalar value per vertex
         """
-        from typing import cast
-
-        # (mypy) numpy stubs often treat np.asarray(...) as Any, so cast explicitly.
-        counts_arr = cast(np.ndarray, np.asarray(face_vertex_counts, dtype=np.int32))
-        indices_arr = cast(np.ndarray, np.asarray(face_vertex_indices, dtype=np.int32))
+        counts_arr = np.asarray(face_vertex_counts, dtype=np.int32)
+        indices_arr = np.asarray(face_vertex_indices, dtype=np.int32)
 
         # Create face ID for each vertex reference
         face_ids = np.repeat(np.arange(len(counts_arr)), counts_arr)
@@ -1284,7 +1450,8 @@ def _uniform_to_vertex_scalar(
 
         # Average
         vertex_scalar = acc / np.maximum(cnt, 1)
-        return cast(np.ndarray, vertex_scalar.astype(np.float32))
+        result: np.ndarray = vertex_scalar.astype(np.float32)
+        return result
 
     def _ensure_vertex_color_material(
         self, stage: Usd.Stage, mesh_prim: Usd.Prim
diff --git a/src/physiomotion4d/vtk_to_usd/__init__.py b/src/physiomotion4d/vtk_to_usd/__init__.py
index e2f88af..82b42c3 100644
--- a/src/physiomotion4d/vtk_to_usd/__init__.py
+++ b/src/physiomotion4d/vtk_to_usd/__init__.py
@@ -30,6 +30,7 @@
 )
 from .usd_mesh_converter import UsdMeshConverter
 from .usd_utils import (
+    add_framing_camera,
     compute_mesh_extent,
     create_primvar,
     ras_normals_to_usd,
@@ -69,6 +70,7 @@
     "sanitize_primvar_name",
     "triangulate_face",
     "compute_mesh_extent",
+    "add_framing_camera",
     # Mesh utils (cell type split)
     "cell_type_name_for_vertex_count",
     "split_mesh_data_by_cell_type",
diff --git a/src/physiomotion4d/vtk_to_usd/converter.py b/src/physiomotion4d/vtk_to_usd/converter.py
index acb3890..5d0c144 100644
--- a/src/physiomotion4d/vtk_to_usd/converter.py
+++ b/src/physiomotion4d/vtk_to_usd/converter.py
@@ -3,11 +3,12 @@
 from pathlib import Path
 from typing import Optional, Union
 
-from pxr import Usd, UsdGeom
+from pxr import Sdf, Usd, UsdGeom
 
 from .data_structures import ConversionSettings, MaterialData
 from .material_manager import MaterialManager
 from .usd_mesh_converter import UsdMeshConverter
+from .usd_utils import add_framing_camera
 from .vtk_reader import read_vtk_file
 
 
@@ -48,6 +49,13 @@ def convert_vtk_file(
     if output_path.exists():
         output_path.unlink()
 
+    # USD caches layers globally by identifier, so a prior call in the same
+    # Python session can block CreateNew even after the file is gone.
+    stale_layer = Sdf.Layer.Find(str(output_path))
+    if stale_layer is not None:
+        stale_layer.Clear()
+        del stale_layer
+
     mesh_data = read_vtk_file(input_path, extract_surface=extract_surface)
 
     stage = Usd.Stage.CreateNew(str(output_path))
@@ -75,5 +83,8 @@ def convert_vtk_file(
     mesh_converter = UsdMeshConverter(stage, conversion_settings, material_mgr)
     mesh_converter.create_mesh(mesh_data, f"{root_path}/{mesh_name}")
 
+    # Framing camera with tight near-clip for Omniverse Kit viewer ergonomics.
+    add_framing_camera(stage)
+
     stage.Save()
     return stage
diff --git a/src/physiomotion4d/vtk_to_usd/material_manager.py b/src/physiomotion4d/vtk_to_usd/material_manager.py
index abdc459..18f6719 100644
--- a/src/physiomotion4d/vtk_to_usd/material_manager.py
+++ b/src/physiomotion4d/vtk_to_usd/material_manager.py
@@ -139,7 +139,10 @@ def bind_material(
             geom_prim: Geometry prim (Mesh, Points, etc.)
             material: Material to bind
         """
-        binding_api = UsdShade.MaterialBindingAPI(geom_prim)
+        # Use Apply() so MaterialBindingAPI ends up in the prim's apiSchemas.
+        # Without this, Omniverse Kit's RTX renderer ignores the binding and
+        # renders the prim as white.
+        binding_api = UsdShade.MaterialBindingAPI.Apply(geom_prim.GetPrim())
         binding_api.Bind(material)
 
         logger.debug(
diff --git a/src/physiomotion4d/vtk_to_usd/primvar_derivations.py b/src/physiomotion4d/vtk_to_usd/primvar_derivations.py
new file mode 100644
index 0000000..c018c83
--- /dev/null
+++ b/src/physiomotion4d/vtk_to_usd/primvar_derivations.py
@@ -0,0 +1,161 @@
+"""Derived primvar computations for VTK-to-USD conversion.
+
+This module hosts a small registry of functions that compute new scalar or
+vector primvars from existing ones, so that downstream colormap workflows can
+visualize physically meaningful quantities (von Mises stress, principal
+invariants, trace, magnitudes, etc.) without depending on the raw tensor
+representation.
+
+Adding a new derivation
+-----------------------
+
+1. Write a function ``derive_<name>(array: GenericArray) -> list[GenericArray]``.
+   It inspects ``array`` (name, num_components, interpolation, data) and returns
+   zero or more derived ``GenericArray`` objects. Returning an empty list means
+   "this derivation does not apply."
+
+2. Append the function to ``PRIMVAR_DERIVATIONS`` below.
+
+The conversion pipeline calls ``derive_primvars`` to apply every registered
+function to every existing primvar; the results are appended to the mesh's
+``generic_arrays`` list.
+
+Naming convention
+-----------------
+
+Use ``<source_name>_<DerivedName>`` (capitalized derived suffix). Capital
+letters sort before lowercase in ASCII, so ``stress_VonMises`` is selected over
+``stress_c0`` by ``USDTools.pick_color_primvar``'s alphabetical tiebreak.
+"""
+
+from __future__ import annotations
+
+import logging
+from typing import Callable
+
+import numpy as np
+from numpy.typing import NDArray
+
+from .data_structures import DataType, GenericArray
+
+logger = logging.getLogger(__name__)
+
+
+# ---------------------------------------------------------------------------
+# Tensor reductions (pure math; no GenericArray plumbing)
+# ---------------------------------------------------------------------------
+
+
+def compute_von_mises_stress(stress_tensor: NDArray) -> NDArray:
+    """Compute scalar von Mises stress from a row-major 9-component tensor.
+
+    The input tensor layout is::
+
+        [s_xx, s_xy, s_xz, s_yx, s_yy, s_yz, s_zx, s_zy, s_zz]
+
+    Off-diagonal pairs are averaged to symmetrize, which is a no-op for an
+    already-symmetric Cauchy stress tensor.
+
+    Formula::
+
+        sigma_VM = sqrt(0.5 * [(sxx-syy)^2 + (syy-szz)^2 + (szz-sxx)^2]
+                        + 3.0 * (sxy^2 + syz^2 + szx^2))
+
+    Args:
+        stress_tensor: Array of shape ``(N, 9)`` or a flat length ``N * 9``.
+
+    Returns:
+        Float32 array of shape ``(N,)`` with per-element von Mises stress.
+    """
+    arr = np.asarray(stress_tensor, dtype=np.float64)
+    if arr.ndim == 1:
+        if arr.size % 9 != 0:
+            raise ValueError(
+                f"Flat stress array length {arr.size} is not divisible by 9"
+            )
+        arr = arr.reshape(-1, 9)
+    elif arr.ndim != 2 or arr.shape[1] != 9:
+        raise ValueError(f"Expected (N, 9) stress tensor, got shape {arr.shape}")
+
+    sxx, sxy, sxz = arr[:, 0], arr[:, 1], arr[:, 2]
+    syx, syy, syz = arr[:, 3], arr[:, 4], arr[:, 5]
+    szx, szy, szz = arr[:, 6], arr[:, 7], arr[:, 8]
+    sym_xy = 0.5 * (sxy + syx)
+    sym_yz = 0.5 * (syz + szy)
+    sym_zx = 0.5 * (sxz + szx)
+
+    deviatoric = 0.5 * ((sxx - syy) ** 2 + (syy - szz) ** 2 + (szz - sxx) ** 2)
+    shear = 3.0 * (sym_xy**2 + sym_yz**2 + sym_zx**2)
+    result: np.ndarray = np.sqrt(np.maximum(deviatoric + shear, 0.0)).astype(np.float32)
+    return result
+
+
+# ---------------------------------------------------------------------------
+# Per-primvar derivation functions
+# ---------------------------------------------------------------------------
+
+
+PrimvarDerivation = Callable[[GenericArray], list[GenericArray]]
+"""Signature for a primvar-derivation function.
+
+A derivation inspects one source ``GenericArray`` and returns zero or more
+newly-computed ``GenericArray`` objects. Return ``[]`` when the derivation
+does not apply to the source array.
+"""
+
+
+def derive_von_mises_from_stress(array: GenericArray) -> list[GenericArray]:
+    """Derive a scalar VonMises primvar from any 9-component stress tensor.
+
+    Matches any source array whose name contains ``"stress"``
+    (case-insensitive) and has 9 components.
+    """
+    if array.num_components != 9 or "stress" not in array.name.lower():
+        return []
+    try:
+        vm = compute_von_mises_stress(array.data)
+    except Exception as exc:
+        logger.debug("Skipping VonMises derivation for '%s': %s", array.name, exc)
+        return []
+    return [
+        GenericArray(
+            name=f"{array.name}_VonMises",
+            data=vm,
+            num_components=1,
+            data_type=DataType.FLOAT,
+            interpolation=array.interpolation,
+        )
+    ]
+
+
+# ---------------------------------------------------------------------------
+# Registry and dispatch
+# ---------------------------------------------------------------------------
+
+
+PRIMVAR_DERIVATIONS: list[PrimvarDerivation] = [
+    derive_von_mises_from_stress,
+]
+"""Ordered list of derivations applied to every input primvar.
+
+Append a function to extend the pipeline. External callers may also mutate
+this list (e.g. to register project-specific derivations at import time), but
+keep in mind the list is module-global.
+"""
+
+
+def derive_primvars(arrays: list[GenericArray]) -> list[GenericArray]:
+    """Apply every registered derivation to every array in ``arrays``.
+
+    Args:
+        arrays: Source primvars. Not modified.
+
+    Returns:
+        The newly-derived primvars only (does not include the originals).
+        Callers typically extend their source list with the result.
+    """
+    derived: list[GenericArray] = []
+    for array in arrays:
+        for deriver in PRIMVAR_DERIVATIONS:
+            derived.extend(deriver(array))
+    return derived
diff --git a/src/physiomotion4d/vtk_to_usd/usd_mesh_converter.py b/src/physiomotion4d/vtk_to_usd/usd_mesh_converter.py
index 9ee0b8e..1f99916 100644
--- a/src/physiomotion4d/vtk_to_usd/usd_mesh_converter.py
+++ b/src/physiomotion4d/vtk_to_usd/usd_mesh_converter.py
@@ -49,6 +49,10 @@ def __init__(
         self.stage = stage
         self.settings = settings
         self.material_mgr = material_mgr
+        # Most recent triangulation face-map from create_mesh(). Reused by
+        # create_time_varying_mesh() when writing per-time-step cell primvars,
+        # since topology (and thus the map) is invariant across time samples.
+        self._last_triangulation_face_map: Optional[np.ndarray] = None
 
     def create_mesh(
         self,
@@ -80,11 +84,17 @@ def create_mesh(
         face_counts = mesh_data.face_vertex_counts
         face_indices = mesh_data.face_vertex_indices
 
+        triangulation_face_map: Optional[np.ndarray] = None
         if self.settings.triangulate_meshes:
             # Check if any faces are not triangles
             if not all(count == 3 for count in face_counts):
                 logger.debug("Triangulating mesh faces")
-                face_counts, face_indices = triangulate_face(face_counts, face_indices)
+                (
+                    face_counts,
+                    face_indices,
+                    triangulation_face_map,
+                ) = triangulate_face(face_counts, face_indices)
+        self._last_triangulation_face_map = triangulation_face_map
 
         # Convert to Vt arrays
         face_counts_vt = Vt.IntArray(face_counts.tolist())
@@ -140,9 +150,12 @@ def create_mesh(
             logger.debug("Adding vertex colors to mesh")
             self._add_vertex_colors(mesh, mesh_data.colors, time_code)
 
-        # Handle generic arrays (primvars)
+        # Handle generic arrays (primvars). Pass the triangulation face-map so
+        # uniform (per-source-face) arrays are expanded to match the
+        # post-triangulation face count; otherwise USD would drop them on size
+        # mismatch.
         if self.settings.preserve_point_arrays or self.settings.preserve_cell_arrays:
-            self._add_generic_arrays(mesh, mesh_data, time_code)
+            self._add_generic_arrays(mesh, mesh_data, time_code, triangulation_face_map)
 
         # Bind material (if material_id is provided and material exists in cache)
         if bind_material and mesh_data.material_id:
@@ -205,7 +218,11 @@ def _add_vertex_colors(
                 display_opacity_primvar.Set(opacity_array)
 
     def _add_generic_arrays(
-        self, mesh: UsdGeom.Mesh, mesh_data: MeshData, time_code: Optional[float]
+        self,
+        mesh: UsdGeom.Mesh,
+        mesh_data: MeshData,
+        time_code: Optional[float],
+        triangulation_face_map: Optional[np.ndarray] = None,
     ) -> None:
         """Add generic data arrays as primvars.
 
@@ -213,8 +230,31 @@ def _add_generic_arrays(
             mesh: USD mesh
             mesh_data: Mesh data containing arrays
             time_code: Optional time code
+            triangulation_face_map: Optional int32 array mapping each
+                triangulated face back to its source face. When provided,
+                uniform-interpolation arrays sized to the source face count
+                are expanded so they match the triangulated face count.
         """
         for array in mesh_data.generic_arrays:
+            if triangulation_face_map is not None and array.interpolation == "uniform":
+                data = np.asarray(array.data)
+                if len(data) == triangulation_face_map.shape[0]:
+                    # Already triangle-aligned (e.g. derived primvar that was
+                    # built post-triangulation). Leave it alone.
+                    pass
+                elif (
+                    len(data) > 0
+                    and triangulation_face_map.size > 0
+                    and triangulation_face_map.max() < len(data)
+                ):
+                    expanded_data = data[triangulation_face_map]
+                    array = GenericArray(
+                        name=array.name,
+                        data=expanded_data,
+                        num_components=array.num_components,
+                        data_type=array.data_type,
+                        interpolation=array.interpolation,
+                    )
             # Avoid authoring large multi-component tensors as flat float[] vertex primvars.
             # Omniverse/Hydra can be unstable when such primvars have elementSize > 1.
             # Instead, split into multiple primvars with <= 3 components each.
@@ -347,12 +387,18 @@ def create_time_varying_mesh(
             if mesh_data.colors is not None:
                 self._add_vertex_colors(mesh, mesh_data.colors, time_code)
 
-            # Update generic arrays
+            # Update generic arrays (reuse the triangulation map computed
+            # for the first time sample; topology is invariant across time).
             if (
                 self.settings.preserve_point_arrays
                 or self.settings.preserve_cell_arrays
             ):
-                self._add_generic_arrays(mesh, mesh_data, time_code)
+                self._add_generic_arrays(
+                    mesh,
+                    mesh_data,
+                    time_code,
+                    self._last_triangulation_face_map,
+                )
 
         logger.info(f"Created time-varying mesh with {len(time_codes)} time samples")
 
diff --git a/src/physiomotion4d/vtk_to_usd/usd_utils.py b/src/physiomotion4d/vtk_to_usd/usd_utils.py
index d93a316..68f23c3 100644
--- a/src/physiomotion4d/vtk_to_usd/usd_utils.py
+++ b/src/physiomotion4d/vtk_to_usd/usd_utils.py
@@ -263,7 +263,7 @@ def create_primvar(
     Args:
         geom: USD geometry prim (Mesh, Points, etc.)
         array: GenericArray containing data
-        array_name_prefix: Prefix for primvar name (e.g., "vtk_point_")
+        array_name_prefix: Prefix for primvar name (e.g., ``"vtk_point_"``)
         time_code: Optional time code for time-varying data
 
     Returns:
@@ -366,44 +366,55 @@ def create_primvar(
     return primvar
 
 
-def triangulate_face(face_counts: NDArray, face_indices: NDArray) -> tuple:
-    """Triangulate polygonal faces.
-
-    Converts quads and polygons to triangles using simple fan triangulation.
+def triangulate_face(
+    face_counts: NDArray, face_indices: NDArray
+) -> tuple[NDArray, NDArray, NDArray]:
+    """Triangulate polygonal faces using fan triangulation.
 
     Args:
-        face_counts: Array of vertex counts per face
-        face_indices: Array of vertex indices
+        face_counts: Array of vertex counts per source face (length F).
+        face_indices: Flat array of vertex indices.
 
     Returns:
-        tuple: (triangulated_counts, triangulated_indices)
+        ``(tri_counts, tri_indices, source_face_index_per_triangle)``:
+        - ``tri_counts``: int32 array, all entries equal to 3.
+        - ``tri_indices``: int32 flat array of triangle vertex indices.
+        - ``source_face_index_per_triangle``: int32 array mapping each output
+          triangle back to its source face in ``face_counts``. Length matches
+          ``tri_counts``. Use this to expand uniform (per-face) primvar data
+          to match the triangulated face count: ``new_data = old_data[mapping]``.
     """
     tri_counts: list[int] = []
     tri_indices: list[int] = []
+    source_face_index: list[int] = []
 
     idx = 0
-    for count in face_counts:
+    for face_idx, count in enumerate(face_counts):
         if count == 3:
-            # Already a triangle
             tri_counts.append(3)
             tri_indices.extend(face_indices[idx : idx + 3])
+            source_face_index.append(face_idx)
         elif count == 4:
-            # Quad -> 2 triangles
             v0, v1, v2, v3 = face_indices[idx : idx + 4]
             tri_counts.extend([3, 3])
             tri_indices.extend([v0, v1, v2, v0, v2, v3])
+            source_face_index.extend([face_idx, face_idx])
         else:
-            # Polygon -> fan triangulation
             v0 = face_indices[idx]
             for i in range(1, count - 1):
                 tri_counts.append(3)
                 tri_indices.extend(
                     [v0, face_indices[idx + i], face_indices[idx + i + 1]]
                 )
+                source_face_index.append(face_idx)
 
         idx += count
 
-    return np.array(tri_counts, dtype=np.int32), np.array(tri_indices, dtype=np.int32)
+    return (
+        np.array(tri_counts, dtype=np.int32),
+        np.array(tri_indices, dtype=np.int32),
+        np.array(source_face_index, dtype=np.int32),
+    )
 
 
 def compute_mesh_extent(points: Vt.Vec3fArray) -> Vt.Vec3fArray:
@@ -416,3 +427,109 @@ def compute_mesh_extent(points: Vt.Vec3fArray) -> Vt.Vec3fArray:
         Vt.Vec3fArray: Extent as [min_point, max_point]
     """
     return UsdGeom.Mesh.ComputeExtent(points)
+
+
+def add_framing_camera(
+    stage: Usd.Stage,
+    *,
+    parent_path: str = "/World",
+    name: str = "Camera",
+    bounds_min: tuple[float, float, float] | None = None,
+    bounds_max: tuple[float, float, float] | None = None,
+    focal_length_mm: float = 50.0,
+    horizontal_aperture_mm: float = 36.0,
+    distance_factor: float = 3.0,
+) -> UsdGeom.Camera | None:
+    """Define a USD camera that frames stage geometry with tight clipping planes.
+
+    Adds a ``UsdGeom.Camera`` prim at ``{parent_path}/{name}`` positioned along
+    +Z to view the supplied (or stage-computed) bounding box. Sets a tight
+    ``clippingRange`` so users can zoom close in Omniverse Kit and other USD
+    viewers without geometry vanishing at the near plane.
+
+    Bounds must be expressed in stage coordinates (post axis-swap and unit
+    scaling). For time-varying stages, bounds are sampled at the start time
+    code.
+
+    Args:
+        stage: The USD stage. Must already contain geometry when bounds are not
+            supplied; world bounds are then computed from the stage.
+        parent_path: Parent prim path. Defaults to ``"/World"``.
+        name: Camera prim name. Defaults to ``"Camera"``.
+        bounds_min: Optional min corner ``(x, y, z)`` in stage coordinates.
+        bounds_max: Optional max corner ``(x, y, z)`` in stage coordinates.
+        focal_length_mm: Camera focal length. USD camera lens parameters are
+            always in millimeters regardless of ``metersPerUnit``.
+        horizontal_aperture_mm: Camera horizontal aperture in millimeters.
+        distance_factor: Camera distance from bbox center as a multiple of the
+            bounding-box diagonal. ``3.0`` gives generous framing.
+
+    Returns:
+        The created Camera prim, or ``None`` if no valid bounds could be found.
+    """
+    if bounds_min is None or bounds_max is None:
+        if stage.HasAuthoredTimeCodeRange():
+            time_code = Usd.TimeCode(stage.GetStartTimeCode())
+        else:
+            time_code = Usd.TimeCode.Default()
+        bbox_cache = UsdGeom.BBoxCache(
+            time_code, includedPurposes=[UsdGeom.Tokens.default_]
+        )
+        world_bbox = bbox_cache.ComputeWorldBound(stage.GetPseudoRoot())
+        bbox_range = world_bbox.ComputeAlignedRange()
+        if bbox_range.IsEmpty():
+            return None
+        bounds_min = tuple(bbox_range.GetMin())
+        bounds_max = tuple(bbox_range.GetMax())
+
+    bmin = np.asarray(bounds_min, dtype=np.float64)
+    bmax = np.asarray(bounds_max, dtype=np.float64)
+    center = 0.5 * (bmin + bmax)
+    size = bmax - bmin
+    diagonal = float(np.linalg.norm(size))
+    if diagonal <= 0.0:
+        return None
+
+    distance = diagonal * distance_factor
+
+    camera_path = f"{parent_path.rstrip('/')}/{name}"
+    camera = UsdGeom.Camera.Define(stage, camera_path)
+
+    # Compute an eye point offset along an axis perpendicular to the stage up
+    # axis, then build a full look-at transform so the camera both moves and
+    # rotates to point at the bbox center. A pure translation only frames the
+    # geometry on a Y-up stage (where the default -Z look direction lines up);
+    # on a Z-up stage it would leave the camera staring horizontally past the
+    # geometry, so we author orientation as well via a single TransformOp.
+    up_axis = UsdGeom.GetStageUpAxis(stage)
+    if up_axis == UsdGeom.Tokens.z:
+        eye = center + np.array([0.0, -distance, 0.0])
+        up_world = Gf.Vec3d(0.0, 0.0, 1.0)
+    else:
+        eye = center + np.array([0.0, 0.0, distance])
+        up_world = Gf.Vec3d(0.0, 1.0, 0.0)
+
+    view = Gf.Matrix4d()
+    view.SetLookAt(
+        Gf.Vec3d(float(eye[0]), float(eye[1]), float(eye[2])),
+        Gf.Vec3d(float(center[0]), float(center[1]), float(center[2])),
+        up_world,
+    )
+    camera_to_world = view.GetInverse()
+
+    # Idempotent: clear any prior xformOpOrder (e.g. a translate op carried in
+    # from a merged source USD that already had a /World/Camera) and author a
+    # single transform op describing the look-at placement.
+    xformable = UsdGeom.Xformable(camera.GetPrim())
+    xformable.ClearXformOpOrder()
+    camera.AddTransformOp().Set(camera_to_world)
+
+    near = max(diagonal * 0.001, 1e-6)
+    far = max(diagonal * 1000.0, distance * 10.0)
+    camera.CreateClippingRangeAttr().Set(Gf.Vec2f(float(near), float(far)))
+
+    camera.CreateFocalLengthAttr().Set(float(focal_length_mm))
+    camera.CreateHorizontalApertureAttr().Set(float(horizontal_aperture_mm))
+    camera.CreateFocusDistanceAttr().Set(float(distance))
+
+    return camera
diff --git a/src/physiomotion4d/workflow_convert_ct_to_vtk.py b/src/physiomotion4d/workflow_convert_ct_to_vtk.py
index 0d207a6..5283ff0 100644
--- a/src/physiomotion4d/workflow_convert_ct_to_vtk.py
+++ b/src/physiomotion4d/workflow_convert_ct_to_vtk.py
@@ -157,13 +157,14 @@ def _create_segmenter(self) -> SegmentAnatomyBase:
     def _get_label_info_for_group(self, group: str) -> tuple[list[str], list[int]]:
         """Return ``(label_names, label_ids)`` for *group* from the active segmenter.
 
-        Reads the segmenter's ``<group>_mask_ids`` dictionary.  Returns empty lists
-        if the attribute does not exist (e.g. simpleware_heart lacks lung/bone ids).
+        Reads the segmenter's :class:`AnatomyTaxonomy`. Returns empty lists if
+        the group is not present (e.g. simpleware_heart does not register
+        lung/bone).
         """
         assert self._segmenter is not None, (
             "_create_segmenter() must be called before _get_label_info_for_group()"
         )
-        mask_ids: dict[int, str] = getattr(self._segmenter, f"{group}_mask_ids", {})
+        mask_ids = self._segmenter.taxonomy.labels_in_group(group)
         return list(mask_ids.values()), list(mask_ids.keys())
 
     @staticmethod
diff --git a/src/physiomotion4d/workflow_convert_heart_gated_ct_to_usd.py b/src/physiomotion4d/workflow_convert_heart_gated_ct_to_usd.py
index 317a2a5..d424775 100644
--- a/src/physiomotion4d/workflow_convert_heart_gated_ct_to_usd.py
+++ b/src/physiomotion4d/workflow_convert_heart_gated_ct_to_usd.py
@@ -7,7 +7,7 @@
 
 import logging
 import os
-from typing import Optional
+from typing import Optional, cast
 
 import itk
 import numpy as np
@@ -43,6 +43,9 @@ def __init__(
         number_of_registration_iterations: Optional[int] = 1,
         registration_method: str = "icon",
         log_level: int | str = logging.INFO,
+        save_registered_images: bool = True,
+        save_registration_transforms: bool = True,
+        save_labelmaps: bool = True,
     ):
         """
         Initialize the Heart-gated CT to USD workflow.
@@ -57,6 +60,12 @@ def __init__(
             number_of_registration_iterations (Optional[int]): Number of registration iterations
             registration_method (str): Registration method to use: 'ants' or 'icon' (default: 'icon')
             log_level: Logging level (default: logging.INFO)
+            save_registered_images: Write registered image intermediates to
+                output_directory when True
+            save_registration_transforms: Write registration transforms to
+                output_directory when True
+            save_labelmaps: Write segmentation labelmaps and registration masks to
+                output_directory when True
         """
         super().__init__(class_name=self.__class__.__name__, log_level=log_level)
 
@@ -66,6 +75,9 @@ def __init__(
         self.project_name = project_name
         self.reference_image_filename = reference_image_filename
         self.number_of_registration_iterations = number_of_registration_iterations
+        self.save_registered_images = save_registered_images
+        self.save_registration_transforms = save_registration_transforms
+        self.save_labelmaps = save_labelmaps
 
         # Validate registration method
         if registration_method not in ["ants", "icon"]:
@@ -123,9 +135,45 @@ def __init__(
         self._time_series_images: list[itk.Image] = []
         self._fixed_image: Optional[itk.Image] = None
         self._fixed_segmentation: Optional[dict[str, itk.Image]] = None
-        self._time_series_transforms: list[itk.Transform] = []
+        self._time_series_transforms: list[dict[str, dict[str, itk.Transform]]] = []
         self._reference_contours: dict[str, pv.PolyData] = {}
 
+    def _output_path(self, filename: str) -> str:
+        """Return an output path inside the workflow output directory."""
+        return os.path.join(self.output_directory, filename)
+
+    def _write_image_if_enabled(
+        self,
+        image: itk.Image,
+        filename: str,
+        enabled: bool,
+    ) -> None:
+        """Write an image artifact when its save option is enabled."""
+        if enabled:
+            itk.imwrite(image, self._output_path(filename), compression=True)
+
+    def _write_transform_if_enabled(
+        self,
+        transform: itk.Transform,
+        filename: str,
+    ) -> None:
+        """Write a transform artifact when transform saving is enabled."""
+        if self.save_registration_transforms:
+            itk.transformwrite(
+                transform,
+                self._output_path(filename),
+                compression=True,
+            )
+
+    def _write_registered_image_if_enabled(self, filename: str) -> None:
+        """Write the current registered moving image when image saving is enabled."""
+        if self.save_registered_images:
+            itk.imwrite(
+                self.registrar.get_registered_image(),
+                self._output_path(filename),
+                compression=True,
+            )
+
     def process(self) -> str:
         """
         Execute the complete workflow from 4D CT to dynamic USD models.
@@ -160,9 +208,8 @@ def _load_time_series(self) -> None:
         if len(self.input_filenames) == 1:
             self.converter.load_nrrd_4d(self.input_filenames[0])
             self.converter.save_3d_images(
-                os.path.join(
-                    self.output_directory, os.path.basename(self.input_filenames[0])
-                )
+                self.output_directory,
+                os.path.basename(self.input_filenames[0]),
             )
         else:
             self.log_info("Loading %d 3D NRRD files", len(self.input_filenames))
@@ -185,10 +232,10 @@ def _load_time_series(self) -> None:
             # Use 70% frame as reference if none specified.
             reference_frame = int(self._num_time_points * 0.7)
             self._fixed_image = self._time_series_images[reference_frame]
-            itk.imwrite(
+            self._write_image_if_enabled(
                 self._fixed_image,
-                os.path.join(self.output_directory, "fixed_image.mha"),
-                compression=True,
+                "fixed_image.mha",
+                self.save_registered_images,
             )
 
         self.log_info("Loaded %d time points", self._num_time_points)
@@ -213,10 +260,10 @@ def _segment_and_register_frames(self) -> None:
         bone_mask = self._fixed_segmentation["bone"]
         other_mask = self._fixed_segmentation["other"]
         contrast_mask = self._fixed_segmentation["contrast"]
-        itk.imwrite(
+        self._write_image_if_enabled(
             labelmap_mask,
-            os.path.join(self.output_directory, "fixed_image_mask.mha"),
-            compression=True,
+            "fixed_image_mask.mha",
+            self.save_labelmaps,
         )
 
         # Create masks for different anatomy types
@@ -254,46 +301,59 @@ def _segment_and_register_frames(self) -> None:
             # Register without mask first
             self.registrar.set_fixed_mask(None)
             result_all = self.registrar.register(moving_image)
-            inverse_transform_all = result_all["inverse_transform"]
-            forward_transform_all = result_all["forward_transform"]
-            itk.transformwrite(
+            inverse_transform_all = cast(itk.Transform, result_all["inverse_transform"])
+            forward_transform_all = cast(itk.Transform, result_all["forward_transform"])
+            self._write_transform_if_enabled(
                 inverse_transform_all,
-                os.path.join(self.output_directory, f"slice_{i:03d}_all_AB.hdf"),
-                compression=True,
+                f"slice_{i:03d}_all_AB.hdf",
             )
-            itk.transformwrite(
+            self._write_transform_if_enabled(
                 forward_transform_all,
-                os.path.join(self.output_directory, f"slice_{i:03d}_all_BA.hdf"),
-                compression=True,
+                f"slice_{i:03d}_all_BA.hdf",
             )
+            self._write_registered_image_if_enabled(f"slice_{i:03d}_registered.mha")
 
-            # Estimate the moving dynamic mask by the inverse transform of the fixed dynamic mask
+            # Estimate the moving dynamic mask from the fixed dynamic mask.
             moving_dynamic_mask = TransformTools().transform_image(
                 fixed_dynamic_mask, inverse_transform_all, moving_image, "nearest"
             )
-            itk.imwrite(
+            self._write_image_if_enabled(
                 moving_dynamic_mask,
-                os.path.join(self.output_directory, f"slice_{i:03d}_dynamic_mask.mha"),
-                compression=True,
+                f"slice_{i:03d}_dynamic_mask.mha",
+                self.save_labelmaps,
             )
             self.registrar.set_fixed_mask(fixed_dynamic_mask)
             result_dynamic = self.registrar.register(moving_image, moving_dynamic_mask)
-            inverse_transform_dynamic = result_dynamic["inverse_transform"]
-            forward_transform_dynamic = result_dynamic["forward_transform"]
+            inverse_transform_dynamic = cast(
+                itk.Transform, result_dynamic["inverse_transform"]
+            )
+            forward_transform_dynamic = cast(
+                itk.Transform, result_dynamic["forward_transform"]
+            )
+            self._write_registered_image_if_enabled(
+                f"slice_{i:03d}_dynamic_registered.mha"
+            )
 
-            # Estimate the moving static mask by the inverse transform of the fixed static mask
+            # Estimate the moving static mask from the fixed static mask.
             moving_static_mask = TransformTools().transform_image(
                 fixed_static_mask, inverse_transform_all, moving_image, "nearest"
             )
-            itk.imwrite(
+            self._write_image_if_enabled(
                 moving_static_mask,
-                os.path.join(self.output_directory, f"slice_{i:03d}_static_mask.mha"),
-                compression=True,
+                f"slice_{i:03d}_static_mask.mha",
+                self.save_labelmaps,
             )
             self.registrar.set_fixed_mask(fixed_static_mask)
             result_static = self.registrar.register(moving_image, moving_static_mask)
-            inverse_transform_static = result_static["inverse_transform"]
-            forward_transform_static = result_static["forward_transform"]
+            inverse_transform_static = cast(
+                itk.Transform, result_static["inverse_transform"]
+            )
+            forward_transform_static = cast(
+                itk.Transform, result_static["forward_transform"]
+            )
+            self._write_registered_image_if_enabled(
+                f"slice_{i:03d}_static_registered.mha"
+            )
 
             # Store transforms
             transforms = {
@@ -310,25 +370,21 @@ def _segment_and_register_frames(self) -> None:
                     "forward_transform": forward_transform_all,
                 },
             }
-            itk.transformwrite(
+            self._write_transform_if_enabled(
                 inverse_transform_dynamic,
-                os.path.join(self.output_directory, f"slice_{i:03d}_dynamic_AB.hdf"),
-                compression=True,
+                f"slice_{i:03d}_dynamic_AB.hdf",
             )
-            itk.transformwrite(
+            self._write_transform_if_enabled(
                 forward_transform_dynamic,
-                os.path.join(self.output_directory, f"slice_{i:03d}_dynamic_BA.hdf"),
-                compression=True,
+                f"slice_{i:03d}_dynamic_BA.hdf",
             )
-            itk.transformwrite(
+            self._write_transform_if_enabled(
                 inverse_transform_static,
-                os.path.join(self.output_directory, f"slice_{i:03d}_static_AB.hdf"),
-                compression=True,
+                f"slice_{i:03d}_static_AB.hdf",
             )
-            itk.transformwrite(
+            self._write_transform_if_enabled(
                 forward_transform_static,
-                os.path.join(self.output_directory, f"slice_{i:03d}_static_BA.hdf"),
-                compression=True,
+                f"slice_{i:03d}_static_BA.hdf",
             )
             self._time_series_transforms.append(transforms)
 
@@ -425,11 +481,14 @@ def _create_usd_files(self) -> None:
         for anatomy_type in ["all", "dynamic", "static"]:
             self.log_info("Creating %s anatomy USD...", anatomy_type)
 
-            # Convert VTK contours to USD
+            # Convert VTK contours to USD. Forwarding the segmenter so labels
+            # land under /World/{project}/{type}/{label_name} (and materials
+            # under /World/Looks/{type}/{label_name}_material).
             converter = ConvertVTKToUSD(
                 self.project_name,
                 self._transformed_contours[anatomy_type],
-                self.segmenter.all_mask_ids,
+                self.segmenter.taxonomy.all_labels(),
+                segmenter=self.segmenter,
                 log_level=self.log_level,
             )
             usd_file = os.path.join(
diff --git a/statistics.md b/statistics.md
index fe67ee9..13b6eb0 100644
--- a/statistics.md
+++ b/statistics.md
@@ -1,471 +1,180 @@
-# PhysioMotion4D - Software Development Statistics & Cost Analysis
+# PhysioMotion4D - Software Development Statistics
 
-**Report Generated:** January 9, 2026  
-**Project Version:** 2025.05.0  
+**Report Generated:** May 13, 2026
+**Project Version:** 2026.05.07
 **Status:** Beta (Development Status: 4 - Beta)
 
 ---
 
 ## Executive Summary
 
-PhysioMotion4D is a sophisticated medical imaging package for generating anatomic models in NVIDIA Omniverse with physiological motion from 4D medical images. This report provides comprehensive statistics on development effort, code quality, and project maturity.
+PhysioMotion4D is a medical imaging package for generating anatomic models in
+NVIDIA Omniverse with physiological motion derived from 4D medical images.
+This report summarizes development effort, code quality, and project maturity.
 
 ### Key Metrics at a Glance
 
-| Metric                         | Value                                 |
-| ------------------------------ | ------------------------------------- |
-| **Total Lines of Code**        | 32,515                                |
-| **Code Coverage**              | 17.64%                                |
-| **Development Period**         | December 5, 2024 - Present (~35 days) |
-| **Active Development Days**    | 6 days                                |
-| **Total Commits**              | 33                                    |
-| **Primary Developer**          | 1 (Stephen Aylward)                   |
-| **Estimated Development Cost** | $125,000 - $175,000 USD               |
-| **Estimated Development Time** | 6-8 person-months                     |
+| Metric                         | Value                                          |
+| ------------------------------ | ---------------------------------------------- |
+| **Total Lines of Code**        | ~42,000                                        |
+| **Development Period**         | December 5, 2025 - May 13, 2026 (~5 months)    |
+| **Total Commits**              | 63                                             |
+| **Primary Developer**          | 1 (Stephen Aylward)                            |
 
 ---
 
-## 📊 Detailed Code Statistics
+## Detailed Code Statistics
 
 ### Lines of Code Breakdown
 
-| Category                 | Files          | Lines of Code | Percentage |
-| ------------------------ | -------------- | ------------- | ---------- |
-| **Core Python Source**   | 28 files       | 13,489        | 41.5%      |
-| **Test Suite**           | 13 files       | 4,506         | 13.9%      |
-| **Jupyter Notebooks**    | 33 files       | 7,295         | 22.4%      |
-| **Documentation**        | ~30 files      | 9,326         | 28.7%      |
-| **Scripts & Automation** | ~5 files       | 127           | 0.4%       |
-| **Infrastructure**       | 4 files        | 1,073         | 3.3%       |
-| **TOTAL**                | **~113 files** | **32,515**    | **100%**   |
-
-### Core Module Breakdown (Python Source)
-
-| Module                                        | Lines | Purpose                                        |
-| --------------------------------------------- | ----- | ---------------------------------------------- |
-| `transform_tools.py`                          | 1,142 | Transform manipulation utilities               |
-| `register_models_pca.py`                      | 818   | PCA-based statistical shape model registration |
-| `workflow_fit_statistical_model_to_patient.py` | 745   | Model-to-patient registration workflow         |
-| `register_images_ants.py`                     | 725   | ANTs-based image registration                  |
-| `segment_anatomy_base.py`                     | 672   | Base class for anatomy segmentation            |
-| `convert_vtk_to_usd_polymesh.py`           | 622   | Polymesh USD conversion                        |
-| `convert_vtk_to_usd_base.py`               | 585   | Base USD conversion functionality              |
-| `workflow_convert_heart_gated_ct_to_usd.py`   | 539   | Heart CT to USD workflow                       |
-| `usd_tools.py`                                | 536   | USD file manipulation                          |
-| `register_time_series_images.py`              | 528   | Time series registration                       |
-| Other modules (18 files)                      | 6,577 | Various specialized functions                  |
-
-### Test Coverage Analysis
-
-```
-Overall Coverage:     17.64%
-Total Valid Lines:    3,708
-Lines Covered:        654
-Lines Not Covered:    3,054
-```
-
-#### Coverage by Module Type
-
-| Module Category                     | Coverage | Status              |
-| ----------------------------------- | -------- | ------------------- |
-| **Base Classes**                    | 62.89%   | ✅ Good              |
-| **Segmentation (TotalSegmentator)** | 91.11%   | ✅ Excellent         |
-| **Segmentation (VISTA-3D)**         | 21.52%   | ⚠️ Needs Improvement |
-| **Image Registration (ANTs)**       | 9.62%    | ⚠️ Needs Improvement |
-| **Image Registration (ICON)**       | 31.37%   | ⚠️ Needs Improvement |
-| **Model Registration (ICP)**        | 16.95%   | ⚠️ Needs Improvement |
-| **Model Registration (PCA)**        | 11.63%   | ⚠️ Needs Improvement |
-| **USD Conversion (Polymesh)**       | 6.20%    | ⚠️ Needs Improvement |
-| **USD Conversion (Tetmesh)**        | 8.19%    | ⚠️ Needs Improvement |
-| **Workflows**                       | 11.60%   | ⚠️ Needs Improvement |
-| **Transform Tools**                 | 9.34%    | ⚠️ Needs Improvement |
-
-**Note:** Low coverage in many modules is typical for early-stage research software focusing on complex medical imaging workflows. The high coverage in base classes and TotalSegmentator indicates mature, well-tested core functionality.
+| Category                             | Files     | Lines of Code | Percentage |
+| ------------------------------------ | --------- | ------------- | ---------- |
+| **Core Python Source (`src/`)**      | 49 files  | 19,670        | 47.3%      |
+| **Test Suite (`tests/`)**            | 24 files  | 7,669         | 18.4%      |
+| **Experiment Scripts (`experiments/`)** | 35 files | 6,687        | 16.1%      |
+| **Tutorial Scripts (`tutorials/`)**  | 10 files  | 1,690         | 4.1%       |
+| **Documentation (`docs/` + repo READMEs)** | ~90 files | ~6,300 rst + ~6,700 md (~13,000 total) | 31%   |
+| **TOTAL**                            | **~210 files** | **~42,000** | **100%**  |
 
----
-
-## 💰 Development Cost Estimation
-
-### COCOMO Model Analysis
-
-Using the Constructive Cost Model (COCOMO) for organic software development:
-
-**Effort Calculation:**
-- **Total Source Lines of Code:** 13,489 (Python) + 7,295 (Notebooks) = 20,784 SLOC
-- **Effective SLOC (excluding comments/blanks, ~70%):** ~14,550 SLOC
-- **COCOMO Effort (Person-Months):** 6.5 PM
-  - Formula: 2.4 × (KSLOC)^1.05 = 2.4 × (14.55)^1.05 ≈ 6.5 PM
+All experiment and tutorial sources are plain `.py` files. Each uses `# %%`
+percent-cell markers so the same file can be executed end-to-end with
+`python <script>.py` or stepped through cell-by-cell in VS Code / Cursor.
 
-**Cost Breakdown:**
+### Core Module Highlights (Python Source)
 
-| Component                    | Effort (PM) | Cost @ $150/hr | Cost @ $200/hr |
-| ---------------------------- | ----------- | -------------- | -------------- |
-| **Core Development**         | 4.0         | $72,000        | $96,000        |
-| **Testing & QA**             | 1.0         | $18,000        | $24,000        |
-| **Documentation**            | 1.0         | $18,000        | $24,000        |
-| **Integration & Deployment** | 0.5         | $9,000         | $12,000        |
-| **Project Management (20%)** | 1.3         | $23,400        | $31,200        |
-| **TOTAL**                    | **7.8 PM**  | **$140,400**   | **$187,200**   |
-
-**Estimated Range:** $125,000 - $175,000 USD (assuming 160 hours/person-month)
-
-### Alternative Calculation (By Component)
-
-| Component           | LOC        | Rate ($/LOC) | Estimated Cost          |
-| ------------------- | ---------- | ------------ | ----------------------- |
-| Core Python Modules | 13,489     | $8-12        | $107,912 - $161,868     |
-| Test Suite          | 4,506      | $5-8         | $22,530 - $36,048       |
-| Documentation       | 9,326      | $2-4         | $18,652 - $37,304       |
-| Notebooks & Scripts | 7,422      | $3-5         | $22,266 - $37,110       |
-| **TOTAL**           | **34,743** | -            | **$171,360 - $272,330** |
-
-**Conservative Estimate:** $150,000 USD  
-**Market Rate Estimate:** $200,000 USD
+| Module                                          | Approx Lines | Purpose                                        |
+| ----------------------------------------------- | ------------ | ---------------------------------------------- |
+| `usd_tools.py`                                  | ~1,500       | USD file manipulation and inspection           |
+| `transform_tools.py`                            | ~1,100       | ITK transform utilities                        |
+| `register_models_pca.py`                        | ~820         | PCA-based shape model registration             |
+| `workflow_fit_statistical_model_to_patient.py`  | ~740         | Model-to-patient registration workflow         |
+| `register_images_ants.py`                       | ~720         | ANTs-based image registration                  |
+| `segment_anatomy_base.py`                       | ~670         | Base class for anatomy segmentation            |
+| `convert_vtk_to_usd.py`                         | ~800         | High-level VTK -> USD converter                |
+| `vtk_to_usd/` subpackage                        | 2,657        | Low-level VTK -> USD building blocks (9 files) |
+| `cli/` subpackage                               | 1,788        | CLI entry-point scripts (8 commands)           |
+| `workflow_convert_heart_gated_ct_to_usd.py`     | ~540         | Heart CT to USD workflow                       |
 
 ---
 
-## 📈 Development Activity & Maturity
-
-### Version Control Statistics
+## Project Maturity Indicators
 
-| Metric                      | Value                        |
-| --------------------------- | ---------------------------- |
-| **Total Commits**           | 33                           |
-| **Active Development Days** | 6                            |
-| **Contributors**            | 1 primary (Stephen Aylward)  |
-| **Project Start Date**      | December 5, 2024             |
-| **Development Duration**    | ~35 days (as of Jan 9, 2026) |
-| **Average Commits per Day** | 5.5 (on active days)         |
-| **Current Version**         | 2025.05.0                    |
-
-### Development Intensity
-
-```
-Commits per Active Day: ████████████████████ 5.5
-Lines per Commit:        ████████████████████ 985
-Files per Commit:        ████████████████████ 3.4
-```
-
-**Analysis:** High productivity per commit suggests mature developer with clear architectural vision. Large LOC per commit indicates significant feature implementations rather than incremental changes.
-
-### Project Maturity Indicators
-
-| Indicator                  | Status                                 | Score               |
-| -------------------------- | -------------------------------------- | ------------------- |
-| **Documentation Coverage** | Comprehensive (28.7% of codebase)      | ✅ Excellent (9/10)  |
-| **Test Suite Present**     | Yes (13.9% of codebase)                | ✅ Good (7/10)       |
-| **Code Coverage**          | 17.64%                                 | ⚠️ Fair (4/10)       |
-| **CI/CD Pipeline**         | GitHub Actions configured              | ✅ Good (7/10)       |
-| **Dependency Management**  | Modern (pyproject.toml, uv)            | ✅ Excellent (9/10)  |
-| **Code Quality Tools**     | Black, Flake8, Pylint, Ruff configured | ✅ Excellent (9/10)  |
-| **Example Notebooks**      | 33 comprehensive examples              | ✅ Excellent (10/10) |
-| **Version Control**        | Structured versioning (bumpver)        | ✅ Good (8/10)       |
-| **API Documentation**      | Docstrings present                     | ✅ Good (7/10)       |
-| **Package Distribution**   | PyPI-ready                             | ✅ Good (8/10)       |
-
-**Overall Maturity Score: 7.8/10** - Beta/Production-Ready
+| Indicator                  | Status                                 |
+| -------------------------- | -------------------------------------- |
+| **Documentation Coverage** | Sphinx site + per-package READMEs      |
+| **Test Suite Present**     | Yes (`tests/` with baselines via Git LFS) |
+| **CI/CD Pipeline**         | GitHub Actions (Ubuntu + Windows; Python 3.11/3.12) |
+| **Dependency Management**  | `pyproject.toml`, `uv`-friendly        |
+| **Code Quality Tools**     | Ruff (lint + format), mypy             |
+| **Example Scripts**        | 35 experiment scripts + 10 tutorial scripts |
+| **Version Management**     | Calendar versioning via bumpver        |
+| **API Reference**          | Google-style docstrings + `docs/API_MAP.md` |
+| **Package Distribution**   | PyPI-ready                             |
 
 ---
 
-## 🏗️ Technical Complexity Assessment
+## Technical Complexity Assessment
 
 ### Domain Complexity
 
-PhysioMotion4D operates in multiple highly complex domains:
+PhysioMotion4D operates across several technically demanding domains:
 
-| Domain                   | Complexity Level | Key Technologies                 |
-| ------------------------ | ---------------- | -------------------------------- |
-| **Medical Imaging**      | Very High        | ITK, MONAI, nibabel              |
-| **Deep Learning**        | High             | PyTorch, CUDA 12.6, transformers |
-| **3D Graphics**          | High             | VTK, PyVista, USD                |
-| **Image Registration**   | Very High        | ANTs, Icon, UniGradICON          |
-| **AI Segmentation**      | High             | TotalSegmentator, VISTA-3D       |
-| **Geometric Processing** | High             | ICP, PCA, distance maps          |
+| Domain                   | Complexity Level | Key Technologies                       |
+| ------------------------ | ---------------- | -------------------------------------- |
+| **Medical Imaging**      | Very High        | ITK, MONAI, nibabel, pynrrd            |
+| **Deep Learning**        | High             | PyTorch, CUDA 13, transformers         |
+| **3D Graphics / USD**    | High             | VTK, PyVista, OpenUSD                  |
+| **Image Registration**   | Very High        | ANTs, Icon, UniGradICON                |
+| **AI Segmentation**      | High             | TotalSegmentator, Simpleware bridge    |
+| **Geometric Processing** | High             | ICP, PCA, distance maps                |
 
 ### Architectural Sophistication
 
-```
-Class Hierarchy Depth:     3-4 levels (well-structured inheritance)
-Module Coupling:           Medium (good separation of concerns)
-Cyclomatic Complexity:     Medium (specialized algorithms)
-Dependency Management:     25+ major dependencies
-```
-
-### Innovation Factors
-
-- **Novel Integration:** First package to bridge 4D medical CT → Omniverse USD
-- **Multi-Modal Registration:** Combines classical (ANTs) and deep learning (Icon) approaches
-- **Ensemble AI:** Multiple segmentation models with intelligent fusion
-- **Physiological Motion:** Captures temporal dynamics of cardiac/respiratory systems
-
-**Complexity Multiplier:** 1.5x (justifies higher development cost estimates)
+- Class hierarchy depth: 3-4 levels (well-structured inheritance from
+  `PhysioMotion4DBase`)
+- Module coupling: medium (clear separation between segmentation,
+  registration, USD conversion, and workflow layers)
+- Public API surface documented in `docs/API_MAP.md`
+- ~25 major external dependencies (medical imaging, AI/ML, USD, registration)
 
 ---
 
-## 📦 Dependencies & Infrastructure
+## Dependencies & Infrastructure
 
-### Core Dependencies (25 Major Packages)
+### Core Dependencies (selected)
 
-| Category              | Count | Key Packages                          |
-| --------------------- | ----- | ------------------------------------- |
-| **Medical Imaging**   | 6     | ITK, TubeTK, nibabel, pynrrd, MONAI   |
-| **Deep Learning**     | 4     | PyTorch, transformers, CUDA libraries |
-| **Registration**      | 3     | ANTs, Icon, UniGradICON               |
-| **3D Graphics**       | 4     | VTK, PyVista, USD-core, trimesh       |
-| **AI Segmentation**   | 2     | TotalSegmentator, VISTA-3D            |
-| **Development Tools** | 10+   | pytest, black, flake8, pylint, sphinx |
+| Category              | Key Packages                                    |
+| --------------------- | ----------------------------------------------- |
+| **Medical Imaging**   | ITK, TubeTK, MONAI, nibabel, pynrrd             |
+| **Deep Learning**     | PyTorch, CuPy (CUDA 13), transformers           |
+| **Registration**      | ANTs, icon-registration, UniGradICON            |
+| **3D Graphics / USD** | VTK, PyVista, USD-core                          |
+| **AI Segmentation**   | TotalSegmentator                                |
+| **Development Tools** | pytest, pytest-xdist, ruff, mypy, sphinx, uv    |
 
 ### Infrastructure Files
 
-| File             | Lines | Purpose                                             |
-| ---------------- | ----- | --------------------------------------------------- |
-| `pyproject.toml` | 368   | Modern Python packaging, dependencies, tool configs |
-| `README.md`      | 467   | Comprehensive project documentation                 |
-| `LICENSE`        | 202   | Apache 2.0 license                                  |
-| `MANIFEST.in`    | 36    | Package distribution manifest                       |
-
-**Total Infrastructure:** 1,073 lines
+| File             | Purpose                                             |
+| ---------------- | --------------------------------------------------- |
+| `pyproject.toml` | Modern Python packaging, dependencies, tool configs |
+| `README.md`      | Repository overview and quick start                 |
+| `LICENSE`        | Apache 2.0 license                                  |
+| `CLAUDE.md`      | Per-repo guidance for Claude Code                   |
+| `AGENTS.md`      | Per-repo guidance for AI coding agents              |
 
 ---
 
-## 🎯 Quality Metrics
+## Quality Metrics
 
 ### Code Quality Configuration
 
-✅ **Black** - Code formatting (line length: 88)  
-✅ **isort** - Import sorting (profile: black)  
-✅ **flake8** - Linting (max line length: 88)  
-✅ **pylint** - Static analysis  
-✅ **ruff** - Fast Python linter  
-✅ **mypy** - Type checking (strict mode)  
-✅ **pre-commit** - Git hooks for quality enforcement
+- **Ruff** - Formatting and linting (line length: 88)
+- **mypy** - Strict type checking (`disallow_untyped_defs = true`)
+- **pre-commit** - Hooks for ruff + mypy + fast tests on push
 
 ### Testing Framework
 
-✅ **pytest** - Testing framework  
-✅ **pytest-cov** - Coverage reporting  
-✅ **pytest-xdist** - Parallel test execution  
-✅ **pytest-timeout** - Timeout control (15 min max)
+- **pytest** - Testing framework
+- **pytest-cov** - Coverage reporting
+- **pytest-xdist** - Parallel test execution
+- **pytest-timeout** - Per-test timeout (15 min default)
 
 **Test Categories:**
 - Unit tests (fast, isolated)
 - Integration tests (slower, multi-component)
 - GPU-dependent tests (segmentation, registration)
 - Data-dependent tests (requires external downloads)
+- Tutorial tests (run each tutorial script end-to-end and compare screenshots)
+- Experiment tests (opt-in via `--run-experiments`; multi-hour run times)
 
 ---
 
-## 📚 Documentation Statistics
-
-### Documentation Distribution
-
-| Type                  | Count       | Lines    | Coverage           |
-| --------------------- | ----------- | -------- | ------------------ |
-| **Markdown Files**    | ~15         | ~5,500   | Comprehensive      |
-| **reStructuredText**  | ~15         | ~3,800   | API + User Guide   |
-| **Jupyter Notebooks** | 33          | 7,295    | Extensive Examples |
-| **Python Docstrings** | All modules | Embedded | Good               |
-| **README**            | 1           | 467      | Excellent          |
-
-### Documentation Quality
-
-- **User Guide:** ✅ Complete with quickstart
-- **API Reference:** ✅ Comprehensive docstrings
-- **Examples:** ✅ 33 tutorial notebooks across 7 categories
-- **Architecture Docs:** ✅ Developer guides present
-- **Testing Guide:** ✅ Comprehensive testing documentation
-- **Contribution Guide:** ✅ Present
-- **FAQ & Troubleshooting:** ✅ Available
-
-**Documentation Score: 9/10** - Professional-grade
-
----
-
-## 🚀 Development Velocity
-
-### Productivity Metrics
-
-```
-Lines of Code per Day (active):     ~5,400 LOC/day
-Files Created per Day (active):     ~19 files/day
-Modules Implemented:                28 core modules
-Test Files Created:                 13 test modules
-Example Notebooks:                  33 notebooks
-```
-
-**Note:** High velocity indicates either:
-1. Experienced developer with domain expertise
-2. Reuse of existing code/patterns
-3. AI-assisted development
-4. Combination of all three
-
-### Feature Completeness
-
-| Feature Category       | Status     | Completeness |
-| ---------------------- | ---------- | ------------ |
-| **Image Segmentation** | ✅ Complete | 100%         |
-| **Image Registration** | ✅ Complete | 100%         |
-| **Model Registration** | ✅ Complete | 100%         |
-| **USD Generation**     | ✅ Complete | 95%          |
-| **Workflows**          | ✅ Complete | 90%          |
-| **CLI Tools**          | ✅ Complete | 80%          |
-| **Documentation**      | ✅ Complete | 95%          |
-| **Test Coverage**      | ⚠️ Partial  | 18%          |
-
-**Overall Feature Completeness: 85%** - Production-ready for core features
-
----
-
-## 📊 Comparative Benchmarks
-
-### Similar Medical Imaging Projects
-
-| Project              | LOC        | Coverage | Team Size | Time      | Status |
-| -------------------- | ---------- | -------- | --------- | --------- | ------ |
-| **PhysioMotion4D**   | 32,515     | 17.64%   | 1         | 1 month   | Beta   |
-| **MONAI Core**       | ~100,000   | 85%+     | 20+       | 3+ years  | Mature |
-| **SimpleITK**        | ~500,000   | 70%+     | 10+       | 10+ years | Mature |
-| **3D Slicer**        | ~1,000,000 | 60%+     | 50+       | 15+ years | Mature |
-| **TotalSegmentator** | ~5,000     | 40%      | 3-5       | 2 years   | Mature |
-
-**Analysis:** PhysioMotion4D achieves significant functionality (32K LOC) in record time (1 month) with minimal resources (1 developer), indicating high efficiency and focused scope.
-
----
-
-## 🎖️ Project Achievements
-
-### Technical Achievements
-
-✅ **Multi-Modal Integration** - Successfully integrates 6+ major medical imaging libraries  
-✅ **AI/ML Pipeline** - Implements state-of-the-art deep learning segmentation  
-✅ **Novel USD Export** - First comprehensive 4D medical → Omniverse bridge  
-✅ **Flexible Architecture** - Extensible base classes for custom implementations  
-✅ **Production-Ready Packaging** - Modern Python packaging with PyPI distribution  
-
-### Code Quality Achievements
-
-✅ **Modern Tooling** - Comprehensive linting, formatting, and type checking  
-✅ **Extensive Examples** - 33 tutorial notebooks covering all features  
-✅ **Professional Documentation** - 9,326 lines of documentation  
-✅ **Test Infrastructure** - Complete test framework with CI/CD  
-✅ **Version Management** - Automated versioning with bumpver  
-
-### Research Impact Potential
-
-- **Medical Visualization:** Enables novel 4D medical visualization in Omniverse
-- **Clinical Applications:** Cardiac and pulmonary motion analysis
-- **Education:** Interactive anatomical models for medical training
-- **Research:** Platform for medical imaging algorithm development
-
----
-
-## 📈 Return on Investment (ROI)
-
-### Value Delivered
-
-| Component               | Market Value | Development Cost | ROI       |
-| ----------------------- | ------------ | ---------------- | --------- |
-| **Core Framework**      | $80,000      | $70,000          | 1.14x     |
-| **AI Integration**      | $50,000      | $30,000          | 1.67x     |
-| **USD/Omniverse**       | $40,000      | $25,000          | 1.60x     |
-| **Documentation**       | $25,000      | $20,000          | 1.25x     |
-| **Test Infrastructure** | $15,000      | $15,000          | 1.00x     |
-| **TOTAL**               | **$210,000** | **$160,000**     | **1.31x** |
+## Documentation Statistics
 
-### Cost Efficiency
+| Type                  | Approx Count       | Approx Lines |
+| --------------------- | ------------------ | ------------ |
+| **Markdown files**    | ~50 (incl. READMEs across `experiments/`, `tests/`, etc.) | ~7,800 |
+| **reStructuredText**  | 71 files under `docs/` | ~6,300 |
+| **Python docstrings** | All public modules | embedded     |
+| **API map**           | 1 generated file (`docs/API_MAP.md`) | ~1,200 |
 
-- **Development Time:** 1 month (vs. typical 6-12 months for similar scope)
-- **Team Size:** 1 developer (vs. typical 3-5 for similar projects)
-- **Quality Level:** Beta/Production-ready in initial release
-- **Time-to-Market:** Exceptional (1 month concept → working software)
+### Documentation Highlights
 
-**Efficiency Multiplier: 6-12x faster than industry standard**
+- Quickstart, tutorials, examples, and architecture under `docs/`
+- Per-subpackage READMEs (e.g. `src/physiomotion4d/vtk_to_usd/CLAUDE.md`)
+- Contribution and testing guides
+- FAQ and troubleshooting sections in test docs
 
 ---
 
-## 🔮 Future Development Recommendations
-
-### Priority 1: Testing & Coverage (High Priority)
-
-- **Target:** Increase coverage from 17.64% to 60%+
-- **Effort:** 2-3 person-months
-- **Cost:** $30,000 - $45,000
-- **Focus Areas:** Registration, USD conversion, workflows
-
-### Priority 2: Performance Optimization (Medium Priority)
-
-- **Target:** 2-3x speedup in registration and segmentation
-- **Effort:** 1-2 person-months
-- **Cost:** $15,000 - $30,000
-- **Focus Areas:** GPU utilization, parallel processing
-
-### Priority 3: Extended Features (Low Priority)
-
-- **Additions:** More anatomical regions, additional AI models
-- **Effort:** 3-4 person-months
-- **Cost:** $45,000 - $60,000
-- **Focus Areas:** Brain, abdomen, vessels
-
-### Priority 4: Cloud Integration (Low Priority)
-
-- **Target:** NVIDIA NIM cloud services, distributed processing
-- **Effort:** 2 person-months
-- **Cost:** $30,000
-- **Focus Areas:** API integration, scalability
-
-**Total Future Investment:** $120,000 - $175,000 (6-11 person-months)
-
----
-
-## 📋 Summary & Conclusion
-
-### Key Findings
-
-1. **Rapid Development:** 32,515 lines of sophisticated medical imaging software in ~35 days
-2. **High Quality:** Professional-grade code with modern tooling and comprehensive documentation
-3. **Cost-Effective:** $150,000-200,000 equivalent value delivered efficiently
-4. **Production-Ready:** Core features complete and functional for beta release
-5. **Technical Excellence:** Successfully integrates complex medical imaging, AI, and 3D graphics domains
-
-### Project Status: **BETA - PRODUCTION READY FOR CORE FEATURES**
-
-### Maturity Assessment
-
-```
-Code Quality:          ████████░░ 8/10
-Documentation:         █████████░ 9/10
-Test Coverage:         ████░░░░░░ 4/10
-Feature Completeness:  ████████░░ 8.5/10
-Architecture:          ████████░░ 8/10
-Performance:           ███████░░░ 7/10
-
-Overall:               ███████░░░ 7.4/10 (Beta/Production-Ready)
-```
-
-### Investment Justification
-
-PhysioMotion4D represents excellent ROI with:
-- **Novel capabilities** in 4D medical → Omniverse pipeline
-- **Professional quality** code and documentation
-- **Rapid development** timeline (6-12x faster than typical)
-- **Extensible architecture** for future growth
-- **Clear path** to production maturity
-
-### Recommended Next Steps
-
-1. ✅ **Release Beta Version** (ready now)
-2. 🎯 **Increase Test Coverage** to 60%+ (2-3 months)
-3. 🚀 **Optimize Performance** for production workloads (1-2 months)
-4. 📈 **Gather User Feedback** from beta users
-5. 🔧 **Iterate Based on Feedback** (ongoing)
-
----
+## Summary
 
-**Report Compiled by:** Automated Analysis System  
-**Data Sources:** Git repository, coverage.xml, pyproject.toml, directory analysis  
-**Analysis Methods:** COCOMO model, industry benchmarks, code metrics  
-**Last Updated:** January 9, 2026
+PhysioMotion4D is a beta-quality medical imaging toolkit that bridges 4D CT
+data, AI segmentation and registration, VTK geometry, and OpenUSD output for
+NVIDIA Omniverse. It is built on top of established medical imaging and 3D
+graphics libraries with a small, focused public API and a percent-cell-script
+example/tutorial layout that runs both interactively and unattended.
 
 ---
 
-*This report represents a snapshot of the PhysioMotion4D project as of January 9, 2026. All cost estimates are based on industry-standard rates and COCOMO modeling for organic software development in the medical imaging domain.*
+**Last Updated:** May 13, 2026
diff --git a/tests/EXPERIMENT_FLAG_USAGE.md b/tests/EXPERIMENT_FLAG_USAGE.md
index e0e6eb1..53ac8de 100644
--- a/tests/EXPERIMENT_FLAG_USAGE.md
+++ b/tests/EXPERIMENT_FLAG_USAGE.md
@@ -4,7 +4,7 @@ This document explains how the `--run-experiments` flag works and why it exists.
 
 ## Purpose
 
-The `--run-experiments` flag provides protection against accidentally running extremely long-running experiment notebook tests. These tests can take 20+ hours to complete and are resource-intensive.
+The `--run-experiments` flag provides protection against accidentally running extremely long-running experiment script tests. These tests can take 20+ hours to complete and are resource-intensive.
 
 ## How It Works
 
@@ -143,7 +143,7 @@ If you ever need to disable this protection (not recommended):
 
 ## Test-Mode Flag: PHYSIOMOTION_RUNNING_AS_TEST
 
-When experiment tests run (with `--run-experiments`), the test runner also sets **`PHYSIOMOTION_RUNNING_AS_TEST=1`** so that notebooks can use reduced parameters (e.g. fewer iterations, fewer files) and finish faster. Notebooks should read this variable and choose quick vs full parameters accordingly. See [EXPERIMENT_TESTS_GUIDE.md](EXPERIMENT_TESTS_GUIDE.md#running-as-test-physiomotion_running_as_test) for the recommended check and the `physiomotion4d.notebook_utils.running_as_test()` helper.
+When experiment tests run (with `--run-experiments`), the test runner also sets **`PHYSIOMOTION_RUNNING_AS_TEST=1`** so that scripts can use reduced parameters (e.g. fewer iterations, fewer files) and finish faster. Scripts should read this variable and choose quick vs full parameters accordingly. See [EXPERIMENT_TESTS_GUIDE.md](EXPERIMENT_TESTS_GUIDE.md#running-as-test-physiomotion_running_as_test) for the recommended check and the `physiomotion4d.test_tools.TestTools.running_as_test()` helper.
 
 ## Related Documentation
 
@@ -153,10 +153,10 @@ When experiment tests run (with `--run-experiments`), the test runner also sets
 
 ## Summary
 
-- ✅ Experiment tests are **opt-in only**
-- ✅ Requires explicit `--run-experiments` flag
-- ✅ Automatically skipped without the flag
-- ✅ Protected from accidental CI/CD execution
-- ✅ Clear, self-documenting behavior
+- Experiment tests are **opt-in only**
+- Requires explicit `--run-experiments` flag
+- Automatically skipped without the flag
+- Protected from accidental CI/CD execution
+- Clear, self-documenting behavior
 
 **Remember:** If you see experiment tests being skipped and want to run them, add `--run-experiments` to your pytest command!
diff --git a/tests/EXPERIMENT_TESTS_GUIDE.md b/tests/EXPERIMENT_TESTS_GUIDE.md
index 7e5cd6c..1cbf847 100644
--- a/tests/EXPERIMENT_TESTS_GUIDE.md
+++ b/tests/EXPERIMENT_TESTS_GUIDE.md
@@ -1,22 +1,27 @@
 # Experiment Tests Guide
 
-This guide explains how to use the automated test suite for PhysioMotion4D experiment notebooks.
+This guide explains how to use the automated test suite for PhysioMotion4D experiment scripts.
 
 ## Overview
 
-The `test_experiments.py` module provides automated testing for all Jupyter notebooks in the `experiments/` directory. Each subdirectory gets its own test that executes all notebooks in alphanumeric order.
+The `test_experiments.py` module provides automated testing for all percent-cell
+Python scripts in the `experiments/` directory. Each subdirectory gets its own
+test that executes every `*.py` script in alphanumeric order.
 
-⚠️ **WARNING:** These tests are **EXTREMELY LONG-RUNNING** and may take multiple hours to complete.
+**WARNING:** These tests are **extremely long-running** and may take multiple
+hours to complete.
 
-🔒 **PROTECTION:** Experiment tests are **opt-in only**. They will NOT run with standard pytest commands like `pytest tests/` or `pytest -v tests/`. You must explicitly pass the `--run-experiments` flag to run them.
+**PROTECTION:** Experiment tests are **opt-in only**. They will NOT run with
+standard pytest commands like `pytest tests/` or `pytest -v tests/`. You must
+explicitly pass the `--run-experiments` flag to run them.
 
 ## Quick Start
 
 ### List Available Experiments
 
 ```bash
-# See all notebooks that would be run (without executing)
-pytest tests/test_experiments.py::test_list_notebooks_in_subdir -v -s --run-experiments
+# See all scripts that would be run (without executing)
+pytest tests/test_experiments.py::test_list_scripts_in_subdir -v -s --run-experiments
 
 # Validate experiment directory structure
 pytest tests/test_experiments.py::test_experiment_structure -v --run-experiments
@@ -32,7 +37,8 @@ pytest tests/test_experiments.py -v --run-experiments
 pytest tests/test_experiments.py -v -m experiment --run-experiments
 ```
 
-⚠️ **IMPORTANT:** Experiment tests require the `--run-experiments` flag to run. Without this flag, they will be automatically skipped.
+**IMPORTANT:** Experiment tests require the `--run-experiments` flag to run.
+Without this flag, they are automatically skipped.
 
 ### Run Individual Experiments
 
@@ -61,41 +67,45 @@ pytest tests/test_experiments.py::test_experiment_heart_statistical_model_to_pat
 # Lung Gated CT to USD (~6 hours, requires GPU & data)
 pytest tests/test_experiments.py::test_experiment_lung_gated_ct_to_usd -v -s --run-experiments
 
-# DISABLED (notebooks not ready):
+# DISABLED (scripts not ready):
 # - test_experiment_displacement_field_to_usd
 # - test_experiment_lung_vessels_airways
 ```
 
-⚠️ **NOTE:** All experiment tests require the `--run-experiments` flag. Without it, they are automatically skipped.
+**NOTE:** All experiment tests require the `--run-experiments` flag. Without
+it, they are automatically skipped.
 
 ## Test Details
 
 ### Available Experiment Tests
 
-| Test Name                                            | Subdirectory                          | Expected Duration | Requirements   | Status         |
-| ---------------------------------------------------- | ------------------------------------- | ----------------- | -------------- | -------------- |
-| `test_experiment_colormap_vtk_to_usd`                | `Colormap-VTK_To_USD/`                | ~2 hours          | Basic          | ✅ Active       |
-| `test_experiment_convert_vtk_to_usd`                 | `Convert_VTK_To_USD/`                 | ~2 hours          | Data           | ✅ Active       |
-| ~~`test_experiment_displacement_field_to_usd`~~      | ~~`DisplacementField_To_USD/`~~       | ~~~2 hours~~      | ~~Basic~~      | 🚫 **Disabled** |
-| `test_experiment_reconstruct_4dct`                   | `Reconstruct4DCT/`                    | ~4 hours          | GPU            | ✅ Active       |
-| `test_experiment_heart_vtk_series_to_usd`            | `Heart-VTKSeries_To_USD/`             | ~3 hours          | Data           | ✅ Active       |
-| `test_experiment_heart_gated_ct_to_usd`              | `Heart-GatedCT_To_USD/`               | ~6 hours          | GPU + Data     | ✅ Active       |
-| `test_experiment_create_statistical_model`           | `Heart-Create_Statistical_Model/`     | ~3 hours          | Data           | ✅ Active       |
-| `test_experiment_heart_statistical_model_to_patient` | `Heart-Statistical_Model_To_Patient/` | ~4 hours          | GPU + Data     | ✅ Active       |
-| `test_experiment_lung_gated_ct_to_usd`               | `Lung-GatedCT_To_USD/`                | ~6 hours          | GPU + Data     | ✅ Active       |
-| ~~`test_experiment_lung_vessels_airways`~~           | ~~`Lung-VesselsAirways/`~~            | ~~~2 hours~~      | ~~GPU + Data~~ | 🚫 **Disabled** |
-
-**Note:** Disabled tests are commented out in the code and will not run. They can be re-enabled when the notebooks are ready.
+| Test Name                                            | Subdirectory                          | Expected Duration | Requirements   | Status     |
+| ---------------------------------------------------- | ------------------------------------- | ----------------- | -------------- | ---------- |
+| `test_experiment_colormap_vtk_to_usd`                | `Colormap-VTK_To_USD/`                | ~2 hours          | Basic          | Active     |
+| `test_experiment_convert_vtk_to_usd`                 | `Convert_VTK_To_USD/`                 | ~2 hours          | Data           | Active     |
+| ~~`test_experiment_displacement_field_to_usd`~~      | ~~`DisplacementField_To_USD/`~~       | ~~~2 hours~~      | ~~Basic~~      | Disabled   |
+| `test_experiment_reconstruct_4dct`                   | `Reconstruct4DCT/`                    | ~4 hours          | GPU            | Active     |
+| `test_experiment_heart_vtk_series_to_usd`            | `Heart-VTKSeries_To_USD/`             | ~3 hours          | Data           | Active     |
+| `test_experiment_heart_gated_ct_to_usd`              | `Heart-GatedCT_To_USD/`               | ~6 hours          | GPU + Data     | Active     |
+| `test_experiment_create_statistical_model`           | `Heart-Create_Statistical_Model/`     | ~3 hours          | Data           | Active     |
+| `test_experiment_heart_statistical_model_to_patient` | `Heart-Statistical_Model_To_Patient/` | ~4 hours          | GPU + Data     | Active     |
+| `test_experiment_lung_gated_ct_to_usd`               | `Lung-GatedCT_To_USD/`                | ~6 hours          | GPU + Data     | Active     |
+| ~~`test_experiment_lung_vessels_airways`~~           | ~~`Lung-VesselsAirways/`~~            | ~~~2 hours~~      | ~~GPU + Data~~ | Disabled   |
+
+**Note:** Disabled tests are commented out in the code and will not run. They
+can be re-enabled when the scripts are ready.
 
 ### Execution Order
 
-Within each subdirectory, notebooks are executed in **alphanumeric order**:
-- `0-download_and_convert_4d_to_3d.ipynb` (runs first)
-- `1-register_images.ipynb` (runs second)
-- `2-generate_segmentation.ipynb` (runs third)
+Within each subdirectory, scripts are executed in **alphanumeric order**
+(the runner uses `sorted(subdir.glob("*.py"))`):
+
+- `0-download_and_convert_4d_to_3d.py` (runs first)
+- `1-register_images.py` (runs second)
+- `2-generate_segmentation.py` (runs third)
 - etc.
 
-This ensures that notebooks with dependencies run in the correct sequence.
+This ensures that scripts with dependencies run in the correct sequence.
 
 ## Requirements
 
@@ -123,7 +133,7 @@ Some experiments require external data downloads:
 - Heart experiments: Slicer-Heart-CT dataset (~1.2GB)
 - Lung experiments: DirLab 4DCT dataset (varies by case)
 
-Data is typically downloaded automatically by the first notebook in each sequence.
+Data is typically downloaded automatically by the first script in each sequence.
 
 ## Test Markers
 
@@ -137,79 +147,92 @@ All experiment tests are marked with:
 
 ## Running as Test: PHYSIOMOTION_RUNNING_AS_TEST
 
-When you run experiment tests with `pytest ... --run-experiments`, the test runner sets the environment variable **`PHYSIOMOTION_RUNNING_AS_TEST=1`** before executing each notebook. Notebooks can read this to use **reduced parameters** (fewer iterations, fewer files, smaller resolution) so test runs complete in reasonable time.
+When you run experiment tests with `pytest ... --run-experiments`, the test
+runner sets the environment variable **`PHYSIOMOTION_RUNNING_AS_TEST=1`** before
+executing each script. Scripts can read this to use **reduced parameters**
+(fewer iterations, fewer files, smaller resolution) so test runs complete in
+reasonable time.
 
 ### How it works
 
-- **Test runner** ([test_experiments.py](test_experiments.py)): `execute_notebook()` passes `env` with `PHYSIOMOTION_RUNNING_AS_TEST=1` to the subprocess that runs `jupyter nbconvert --execute`, so the notebook kernel sees the variable.
-- **Notebooks**: In an early cell (e.g. after imports or with other config), compute a boolean and use it to choose quick vs full parameters.
+- **Test runner** ([test_experiments.py](test_experiments.py)): `execute_script()`
+  passes `env` with `PHYSIOMOTION_RUNNING_AS_TEST=1` to the subprocess that runs
+  the script as a standalone Python file, so the script process sees the variable.
+- **Scripts**: Early in the script (after imports or with other config),
+  compute a boolean and use it to choose quick vs full parameters.
 
-### Recommended check in notebooks
+### Recommended check in scripts
 
 Use either of these:
 
-**Option 1 – inline (no extra import):**
+**Option 1 - inline (stdlib only):**
 
 ```python
+import os
+
 running_as_test = os.environ.get("PHYSIOMOTION_RUNNING_AS_TEST", "").lower() in ("1", "true", "yes")
 ```
 
-**Option 2 – shared helper (recommended):**
+**Option 2 - shared helper (recommended):**
 
 ```python
-from physiomotion4d.notebook_utils import running_as_test
+from physiomotion4d.test_tools import TestTools
 
-# Then use running_as_test() where you need it, e.g.:
-quick_run = running_as_test()
-max_iterations = 100 if running_as_test() else 2000
+# Then use TestTools.running_as_test() where you need it, e.g.:
+quick_run = TestTools.running_as_test()
+max_iterations = 100 if TestTools.running_as_test() else 2000
 ```
 
 ### Semantics
 
-- **Truthy values** (case-insensitive): `1`, `true`, `yes` → notebook should use fast/small parameters.
+- **Truthy values** (case-insensitive): `1`, `true`, `yes` -> script should use
+  fast/small parameters.
 - **Unset or falsy**: use full parameters (normal interactive or production run).
 
-Notebooks that support this will run quickly when executed as tests and at full fidelity when run manually.
+Scripts that support this will run quickly when executed as tests and at full
+fidelity when run manually.
 
 ## Usage Tips
 
 ### Run with Detailed Output
 
 ```bash
-# Show all output from notebooks (recommended)
+# Show all output from scripts (recommended)
 pytest tests/test_experiments.py::test_experiment_heart_gated_ct_to_usd -v -s
 ```
 
 The `-s` flag shows all stdout/stderr, including:
-- Notebook execution progress
-- Cell outputs
+- Script execution progress
+- Cell outputs (`# %%` percent-cell scripts are run as ordinary Python)
 - Error messages
 - Execution summaries
 
 ### Monitor Progress
 
-Each notebook execution prints:
+Each script execution prints:
 ```
 ================================================================================
-Executing notebook: 1-register_images.ipynb
-Path: experiments/Heart-GatedCT_To_USD/1-register_images.ipynb
+Executing script: 1-register_images.py
+Path: experiments/Heart-GatedCT_To_USD/1-register_images.py
 Timeout: 5400 seconds (90 minutes)
 ================================================================================
 
-... (notebook output) ...
+... (script output) ...
 
-✅ Successfully executed: 1-register_images.ipynb
+Successfully executed: 1-register_images.py
 ```
 
 ### Handle Failures
 
-If a notebook fails:
+If a script fails:
 1. Check the error output in the pytest summary
-2. Open the notebook file to see execution results
+2. Open the script to inspect the failing cell
 3. Fix the issue (code, data, environment)
 4. Re-run the specific test
 
-The notebooks are executed **in place** (with `--inplace` flag), so execution results are saved in the notebook file.
+The scripts are executed as ordinary Python (`python <script>.py`); cell
+boundaries from `# %%` markers are only meaningful inside cell-aware editors
+and do not affect test execution.
 
 ### Skip Tests by Marker
 
@@ -224,13 +247,15 @@ pytest tests/test_experiments.py -m "experiment and not requires_data" -v --run-
 pytest tests/test_experiments.py -m "experiment and not requires_gpu and not requires_data" -v --run-experiments
 ```
 
-**Note:** The `--run-experiments` flag is always required, regardless of marker filters.
+**Note:** The `--run-experiments` flag is always required, regardless of marker
+filters.
 
 ## CI/CD Exclusion
 
 **These tests are NEVER run in CI/CD workflows.**
 
-Experiment tests are protected by requiring the `--run-experiments` flag, which is never used in CI/CD. Even if someone accidentally runs:
+Experiment tests are protected by requiring the `--run-experiments` flag, which
+is never used in CI/CD. Even if someone accidentally runs:
 
 ```bash
 # These commands will NOT run experiment tests (they'll be skipped)
@@ -270,7 +295,7 @@ If a test times out:
 
 If you get OOM errors:
 1. Close other applications
-2. Reduce batch sizes in notebooks (if applicable)
+2. Reduce batch sizes in scripts (if applicable)
 3. Use smaller test datasets
 4. Consider running on a machine with more RAM
 
@@ -286,15 +311,15 @@ If GPU tests fail due to CUDA issues:
 
 If data downloads fail:
 1. Check internet connection
-2. Verify URLs in notebooks are still valid
+2. Verify URLs in scripts are still valid
 3. Manually download data and place in expected location
 4. Check disk space
 
-### Notebook Execution Errors
+### Script Execution Errors
 
-If individual cells fail:
-1. Open the notebook in JupyterLab
-2. Run cells manually to identify the issue
+If a script fails:
+1. Open the script in your editor (VS Code / Cursor with the Python extension)
+2. Run the cells in order to identify the issue
 3. Check for:
    - Missing dependencies
    - Incorrect paths
@@ -305,11 +330,11 @@ If individual cells fail:
 
 ### Customizing Timeouts
 
-Edit `test_experiments.py` to adjust per-notebook or per-test timeouts:
+Edit `test_experiments.py` to adjust per-script or per-test timeouts:
 
 ```python
-# Per-notebook timeout (default: 3600 seconds = 1 hour)
-run_experiment_notebooks('Heart-GatedCT_To_USD', timeout_per_notebook=7200)
+# Per-script timeout (default: 3600 seconds = 1 hour)
+run_experiment_scripts('Heart-GatedCT_To_USD', timeout_per_script=7200)
 
 # Per-test timeout decorator
 @pytest.mark.timeout(21600)  # 6 hours
@@ -322,7 +347,7 @@ def test_experiment_...():
 To add a new experiment directory:
 
 1. Create the subdirectory in `experiments/`
-2. Add notebooks (use numeric prefixes for ordering)
+2. Add `# %%` percent-cell Python scripts (use numeric prefixes for ordering)
 3. Add to `EXPERIMENT_SUBDIRS` list in `test_experiments.py`:
    ```python
    EXPERIMENT_SUBDIRS = [
@@ -337,13 +362,13 @@ To add a new experiment directory:
    @pytest.mark.slow
    @pytest.mark.timeout(7200)
    def test_experiment_your_new_experiment():
-       """Test YourNewExperiment notebooks."""
-       run_experiment_notebooks('YourNewExperiment', timeout_per_notebook=3600)
+       """Test YourNewExperiment scripts."""
+       run_experiment_scripts('YourNewExperiment', timeout_per_script=3600)
    ```
 
 ### Parallel Execution
 
-⚠️ **Use with caution** - Experiments use significant GPU/CPU resources.
+**Use with caution** - Experiments use significant GPU/CPU resources.
 
 You can run multiple experiment subdirectories in parallel using pytest-xdist:
 
@@ -356,11 +381,11 @@ pytest tests/test_experiments.py -v -n auto --run-experiments
 ```
 
 **IMPORTANT GUARANTEES:**
-- ✅ Notebooks **within** each subdirectory run **SEQUENTIALLY** in alphanumeric order
-- ✅ Different subdirectories **CAN** run in parallel (if resources allow)
-- ✅ Each test function is atomic - pytest-xdist won't split notebook execution
-- ✅ `@pytest.mark.xdist_group` markers prevent conflicts if needed
-- ✅ Execution stops on first failure within each subdirectory
+- Scripts **within** each subdirectory run **SEQUENTIALLY** in alphanumeric order
+- Different subdirectories **CAN** run in parallel (if resources allow)
+- Each test function is atomic - pytest-xdist won't split script execution
+- `@pytest.mark.xdist_group` markers prevent conflicts if needed
+- Execution stops on first failure within each subdirectory
 
 **Example:**
 ```bash
@@ -368,33 +393,38 @@ pytest tests/test_experiments.py -v -n auto --run-experiments
 # Worker 1: test_experiment_colormap_vtk_to_usd
 # Worker 2: test_experiment_heart_vtk_series_to_usd
 
-# But within Worker 1's test, notebooks run in strict order:
-# 1. colormap_vtk_to_usd.ipynb (must complete first)
-# ... (any subsequent notebooks in that directory)
+# But within Worker 1's test, scripts run in strict order:
+# 1. colormap_vtk_to_usd.py (must complete first)
+# ... (any subsequent scripts in that directory)
 ```
 
 ## FAQ
 
 **Q: How long will all experiments take?**
-A: Approximately 32 hours total on a high-end workstation (GPU, 64GB RAM). Note: DisplacementField_To_USD and Lung-VesselsAirways are currently disabled.
+A: Approximately 32 hours total on a high-end workstation (GPU, 64GB RAM). Note:
+DisplacementField_To_USD and Lung-VesselsAirways are currently disabled.
 
 **Q: Can I run these on CPU only?**
-A: Some experiments will work, but most require GPU and will be extremely slow or fail on CPU.
-
-**Q: Will these tests modify the notebooks?**
-A: Yes, notebooks are executed in-place. Execution results are saved in the notebook files.
+A: Some experiments will work, but most require GPU and will be extremely slow
+or fail on CPU.
 
-**Q: Should I commit notebook outputs?**
-A: Depends on your workflow. Consider using `nbstripout` to clean outputs before committing.
+**Q: Will these tests modify the scripts?**
+A: No, scripts are executed as ordinary Python processes. The scripts may
+write outputs to disk (per their own logic) but the script files themselves are
+not modified.
 
-**Q: What if I only want to test specific notebooks?**
-A: The test suite runs all notebooks in a subdirectory. To test individual notebooks, run them manually in JupyterLab.
+**Q: What if I only want to test specific scripts?**
+A: The test suite runs all scripts in a subdirectory. To test individual
+scripts, run them manually with `python path/to/script.py` or use VS
+Code/Cursor's "Run Cell" support for percent-cell scripts.
 
 **Q: Can I use these tests for regression testing?**
-A: Yes, but you'll need to add output validation. Currently, tests only check that notebooks execute without errors.
+A: Yes, but you'll need to add output validation. Currently, tests only check
+that scripts execute without errors.
 
 **Q: Are these tests deterministic?**
-A: Not always - some experiments use random seeds, AI models, or external data that may change.
+A: Not always - some experiments use random seeds, AI models, or external data
+that may change.
 
 ## Related Documentation
 
@@ -421,7 +451,7 @@ If you encounter issues:
 
 1. Check this guide and related documentation
 2. Review the test output carefully
-3. Try running the notebook manually in JupyterLab
+3. Try running the script manually with `python path/to/script.py`
 4. Check GitHub issues for similar problems
 5. Open a new issue with:
    - Test command used
@@ -431,4 +461,5 @@ If you encounter issues:
 
 ---
 
-**Remember:** These tests are for validation and regression testing. They are NOT part of CI/CD and must be run manually.
+**Remember:** These tests are for validation and regression testing. They are
+NOT part of CI/CD and must be run manually.
diff --git a/tests/EXPERIMENT_TESTS_SUMMARY.md b/tests/EXPERIMENT_TESTS_SUMMARY.md
index 2dfe029..1bf7577 100644
--- a/tests/EXPERIMENT_TESTS_SUMMARY.md
+++ b/tests/EXPERIMENT_TESTS_SUMMARY.md
@@ -2,18 +2,21 @@
 
 ## Overview
 
-This document summarizes the implementation of automated testing for PhysioMotion4D experiment notebooks.
+This document summarizes the implementation of automated testing for
+PhysioMotion4D experiment scripts.
 
 **Created:** 2026-01-25
 
 ## What Was Implemented
 
-A comprehensive test suite (`test_experiments.py`) that automatically executes all Jupyter notebooks in the `experiments/` directory, organized by subdirectory. Each subdirectory gets its own dedicated test function.
+A comprehensive test suite (`test_experiments.py`) that automatically executes
+all percent-cell Python scripts in the `experiments/` directory, organized by
+subdirectory. Each subdirectory gets its own dedicated test function.
 
 ### Key Features
 
-1. **Automated notebook execution** - Uses `nbconvert` to execute notebooks in place
-2. **Alphanumeric ordering** - Notebooks run in order (0-, 1-, 2-, etc.)
+1. **Automated script execution** - Runs each `*.py` script as a subprocess
+2. **Alphanumeric ordering** - Scripts run in order (0-, 1-, 2-, etc.)
 3. **One test per subdirectory** - Each experiment gets its own test function
 4. **Long timeouts** - Tests accommodate hours-long execution times
 5. **Detailed reporting** - Progress updates, summaries, and failure diagnostics
@@ -22,38 +25,38 @@ A comprehensive test suite (`test_experiments.py`) that automatically executes a
 
 ### Covered Experiment Subdirectories
 
-1. **Colormap-VTK_To_USD** - VTK colormap conversion to USD ✅ Active
-2. **Convert_VTK_To_USD** - VTK to USD using library classes ✅ Active
-3. ~~**DisplacementField_To_USD**~~ - Registration displacement field visualization 🚫 **Disabled (notebooks not ready)**
-4. **Reconstruct4DCT** - 4D CT reconstruction techniques ✅ Active
-5. **Heart-VTKSeries_To_USD** - Heart VTK time series conversion ✅ Active
-6. **Heart-GatedCT_To_USD** - Complete cardiac imaging pipeline ✅ Active
-7. **Heart-Create_Statistical_Model** - PCA statistical shape model creation ✅ Active
-8. **Heart-Statistical_Model_To_Patient** - Heart model to patient registration with PCA ✅ Active
-9. **Lung-GatedCT_To_USD** - Lung imaging pipeline with DirLab data ✅ Active
-10. ~~**Lung-VesselsAirways**~~ - Vessel and airway segmentation 🚫 **Disabled (notebooks not ready)**
+1. **Colormap-VTK_To_USD** - VTK colormap conversion to USD - Active
+2. **Convert_VTK_To_USD** - VTK to USD using library classes - Active
+3. ~~**DisplacementField_To_USD**~~ - Registration displacement field visualization - **Disabled (scripts not ready)**
+4. **Reconstruct4DCT** - 4D CT reconstruction techniques - Active
+5. **Heart-VTKSeries_To_USD** - Heart VTK time series conversion - Active
+6. **Heart-GatedCT_To_USD** - Complete cardiac imaging pipeline - Active
+7. **Heart-Create_Statistical_Model** - PCA statistical shape model creation - Active
+8. **Heart-Statistical_Model_To_Patient** - Heart model to patient registration with PCA - Active
+9. **Lung-GatedCT_To_USD** - Lung imaging pipeline with DirLab data - Active
+10. ~~**Lung-VesselsAirways**~~ - Vessel and airway segmentation - **Disabled (scripts not ready)**
 
 **8 active experiments, 2 disabled**
 
 ## Files Created
 
-### 1. `tests/test_experiments.py` (435 lines)
+### 1. `tests/test_experiments.py`
 
 Main test module containing:
-- Notebook discovery and execution functions
+- Script discovery and execution functions
 - One test function per experiment subdirectory
-- Helper tests for listing notebooks and validating structure
+- Helper tests for listing scripts and validating structure
 - Comprehensive docstrings with usage examples
 
 **Key functions:**
-- `get_notebooks_in_subdir()` - Find notebooks in a subdirectory
-- `execute_notebook()` - Execute a single notebook with timeout
-- `run_experiment_notebooks()` - Execute all notebooks in a subdirectory
+- `get_scripts_in_subdir()` - Find scripts in a subdirectory
+- `execute_script()` - Execute a single script with timeout
+- `run_experiment_scripts()` - Execute all scripts in a subdirectory
 - `test_experiment_*()` - Individual test functions (8 total)
 - `test_experiment_structure()` - Validate experiment directory structure
-- `test_list_notebooks_in_subdir()` - Helper to preview notebooks
+- `test_list_scripts_in_subdir()` - Helper to preview scripts
 
-### 2. `tests/EXPERIMENT_TESTS_GUIDE.md` (480 lines)
+### 2. `tests/EXPERIMENT_TESTS_GUIDE.md`
 
 Comprehensive usage guide containing:
 - Quick start examples
@@ -75,7 +78,7 @@ Added new pytest marker:
 ```python
 markers = [
     # ... existing markers ...
-    "experiment: marks tests that run experiment notebooks (extremely slow, manual only)"
+    "experiment: marks tests that run experiment scripts (extremely slow, manual only)"
 ]
 ```
 
@@ -120,43 +123,43 @@ Added **Automated Testing** section with:
 ### Run All Experiments
 
 ```bash
-pytest tests/test_experiments.py -v -m experiment
+pytest tests/test_experiments.py -v -m experiment --run-experiments
 ```
 
 ### Run Single Experiment
 
 ```bash
-pytest tests/test_experiments.py::test_experiment_heart_gated_ct_to_usd -v -s
+pytest tests/test_experiments.py::test_experiment_heart_gated_ct_to_usd -v -s --run-experiments
 ```
 
-### List Notebooks (No Execution)
+### List Scripts (No Execution)
 
 ```bash
-pytest tests/test_experiments.py::test_list_notebooks_in_subdir -v -s
+pytest tests/test_experiments.py::test_list_scripts_in_subdir -v -s --run-experiments
 ```
 
 ### Validate Structure
 
 ```bash
-pytest tests/test_experiments.py::test_experiment_structure -v
+pytest tests/test_experiments.py::test_experiment_structure -v --run-experiments
 ```
 
 ## Test Characteristics
 
 ### Timeouts
 
-| Test                               | Per-Notebook   | Total Test     | Status     |
+| Test                               | Per-Script     | Total Test     | Status     |
 | ---------------------------------- | -------------- | -------------- | ---------- |
-| Colormap VTK to USD                | 3600s (1h)     | 7200s (2h)     | ✅ Active   |
-| Convert VTK to USD                 | 3600s (1h)     | 7200s (2h)     | ✅ Active   |
-| ~~Displacement Field~~             | ~~3600s (1h)~~ | ~~7200s (2h)~~ | 🚫 Disabled |
-| Reconstruct 4DCT                   | 7200s (2h)     | 14400s (4h)    | ✅ Active   |
-| Heart VTK Series                   | 5400s (1.5h)   | 10800s (3h)    | ✅ Active   |
-| Heart Gated CT                     | 5400s (1.5h)   | 21600s (6h)    | ✅ Active   |
-| Create Statistical Model           | 5400s (1.5h)   | 10800s (3h)    | ✅ Active   |
-| Heart Statistical Model to Patient | 7200s (2h)     | 14400s (4h)    | ✅ Active   |
-| Lung Gated CT                      | 5400s (1.5h)   | 21600s (6h)    | ✅ Active   |
-| ~~Lung Vessels Airways~~           | ~~3600s (1h)~~ | ~~7200s (2h)~~ | 🚫 Disabled |
+| Colormap VTK to USD                | 3600s (1h)     | 7200s (2h)     | Active     |
+| Convert VTK to USD                 | 3600s (1h)     | 7200s (2h)     | Active     |
+| ~~Displacement Field~~             | ~~3600s (1h)~~ | ~~7200s (2h)~~ | Disabled   |
+| Reconstruct 4DCT                   | 7200s (2h)     | 14400s (4h)    | Active     |
+| Heart VTK Series                   | 5400s (1.5h)   | 10800s (3h)    | Active     |
+| Heart Gated CT                     | 5400s (1.5h)   | 21600s (6h)    | Active     |
+| Create Statistical Model           | 5400s (1.5h)   | 10800s (3h)    | Active     |
+| Heart Statistical Model to Patient | 7200s (2h)     | 14400s (4h)    | Active     |
+| Lung Gated CT                      | 5400s (1.5h)   | 21600s (6h)    | Active     |
+| ~~Lung Vessels Airways~~           | ~~3600s (1h)~~ | ~~7200s (2h)~~ | Disabled   |
 
 **Total for active experiments: ~32 hours** (2 experiments disabled)
 
@@ -180,20 +183,21 @@ All GitHub Actions workflows explicitly exclude them:
 
 ## Design Decisions
 
-### 1. Execution Method: nbconvert
+### 1. Execution Method: subprocess
 
-**Chosen:** `nbconvert --execute --inplace`
+**Chosen:** Run each script as a normal Python subprocess (`python <script>.py`).
 
 **Why:**
-- Standard tool, comes with JupyterLab
-- Executes notebooks with real kernel
-- Preserves outputs in notebook file
-- Good timeout and error handling
+- Scripts are plain Python files with `# %%` cell markers; they execute
+  end-to-end without any cell-aware runtime.
+- Clean process boundary makes timeouts and environment variables easy to manage.
+- No extra dependencies beyond the standard library.
 
 **Alternatives considered:**
-- papermill - More features, but extra dependency
-- nbclient - Lower level, more complex
-- Manual kernel execution - Too complex
+- An in-process exec of each script file - simpler but would leak module-level
+  state across scripts.
+- `runpy.run_path(..., run_name="__main__")` - works for tutorials but not for
+  multi-hour scripts that may need their own process for cleanup.
 
 ### 2. One Test Per Subdirectory
 
@@ -207,14 +211,15 @@ All GitHub Actions workflows explicitly exclude them:
 - Pros: Less code duplication
 - Cons: Harder to run specific experiments, less clear output
 
-### 3. In-Place Execution
+### 3. Out-of-Process Execution
 
 **Why:**
-- Notebooks retain execution results
-- Easier debugging (open notebook to see outputs)
-- Matches interactive development workflow
+- Each script gets a clean interpreter (no leaked global state across scripts).
+- Failures isolate to the failing script.
+- Subprocess stdout/stderr can be captured cleanly by pytest's `-s` flag.
 
-**Downside:** Modifies notebooks (consider using nbstripout)
+**Downside:** Slower per-script startup, but negligible compared to multi-hour
+runtimes.
 
 ### 4. Alphanumeric Ordering
 
@@ -223,7 +228,7 @@ All GitHub Actions workflows explicitly exclude them:
 - Matches dependency order in experiments
 - Intuitive and predictable
 
-**Implementation:** `sorted(subdir.glob('*.ipynb'))`
+**Implementation:** `sorted(subdir.glob('*.py'))`
 
 ### 5. Long Timeouts
 
@@ -233,7 +238,7 @@ All GitHub Actions workflows explicitly exclude them:
 - Data downloads can take time
 - Better to have generous timeouts than false failures
 
-**Trade-off:** Hung notebooks take long to fail
+**Trade-off:** Hung scripts take long to fail.
 
 ## Testing the Tests
 
@@ -241,17 +246,17 @@ To verify the implementation works:
 
 1. **Structure validation:**
    ```bash
-   pytest tests/test_experiments.py::test_experiment_structure -v
+   pytest tests/test_experiments.py::test_experiment_structure -v --run-experiments
    ```
 
-2. **Notebook discovery:**
+2. **Script discovery:**
    ```bash
-   pytest tests/test_experiments.py::test_list_notebooks_in_subdir -v -s
+   pytest tests/test_experiments.py::test_list_scripts_in_subdir -v -s --run-experiments
    ```
 
 3. **Run fastest experiment:**
    ```bash
-   pytest tests/test_experiments.py::test_experiment_colormap_vtk_to_usd -v -s
+   pytest tests/test_experiments.py::test_experiment_colormap_vtk_to_usd -v -s --run-experiments
    ```
 
 4. **Verify CI/CD exclusion:**
@@ -265,14 +270,14 @@ To verify the implementation works:
 Potential improvements for future iterations:
 
 1. **Parallel execution** - Run independent experiments simultaneously
-2. **HTML reports** - Generate browsable report from executed notebooks
+2. **HTML reports** - Generate browsable report from executed scripts
 3. **Artifact collection** - Automatically gather generated USD files, images
-4. **Smoke tests** - Quick mode that only runs first cell of each notebook
-5. **Parameterization** - Use papermill to parameterize notebook execution
-6. **Checkpointing** - Resume from failed notebook in sequence
+4. **Smoke tests** - Quick mode that only runs the first cell of each script
+5. **Parameterization** - Inject parameters via env vars or CLI args
+6. **Checkpointing** - Resume from failed script in sequence
 7. **Resource monitoring** - Track memory, GPU, disk usage during execution
 8. **Baseline comparison** - Compare outputs with known-good results
-9. **Partial execution** - Run only specific notebooks within a subdirectory
+9. **Partial execution** - Run only specific scripts within a subdirectory
 10. **Cleanup hooks** - Automatically remove temporary files after tests
 
 ## Maintenance Notes
@@ -284,24 +289,24 @@ When adding a new experiment subdirectory:
 1. Add to `EXPERIMENT_SUBDIRS` list in `test_experiments.py`
 2. Create a test function following the naming pattern
 3. Choose appropriate markers and timeout
-4. Document expected notebooks in docstring
+4. Document expected scripts in docstring
 5. Update this summary and the guide
 
 ### Modifying Timeouts
 
 If experiments become faster/slower:
 
-1. Update `timeout_per_notebook` argument
+1. Update `timeout_per_script` argument
 2. Update `@pytest.mark.timeout()` decorator
 3. Update documentation (guide and summary)
 
-### Handling Notebook Changes
+### Handling Script Changes
 
-If notebook dependencies or order changes:
+If script dependencies or order changes:
 
 - Tests automatically use alphanumeric ordering
-- No code changes needed unless notebooks are added/removed
-- Consider updating test docstrings if notebook purposes change
+- No code changes needed unless scripts are added/removed
+- Consider updating test docstrings if script purposes change
 
 ## Known Limitations
 
@@ -310,7 +315,7 @@ If notebook dependencies or order changes:
 3. **Resource intensive** - Requires significant compute, GPU, memory, disk
 4. **Long runtime** - Hours to complete all tests
 5. **External dependencies** - Requires data downloads, internet connectivity
-6. **Modifies files** - Notebooks executed in-place with outputs saved
+6. **Side effects on disk** - Scripts write outputs to disk by design
 7. **Limited parallelism** - Tests run sequentially by default
 8. **No cleanup** - Generated files not automatically removed
 
@@ -326,40 +331,41 @@ If notebook dependencies or order changes:
 
 For issues:
 1. Consult `EXPERIMENT_TESTS_GUIDE.md`
-2. Check test output and notebook execution results
+2. Check test output and script execution results
 3. Verify system requirements (GPU, memory, disk)
-4. Try running notebook manually in JupyterLab
-5. Check GitHub issues or open new one
+4. Try running the script manually with `python path/to/script.py`
+5. Check GitHub issues or open a new one
 
 ## Conclusion
 
-This implementation provides a solid foundation for automated validation of PhysioMotion4D experiment notebooks. The test suite is:
+This implementation provides a solid foundation for automated validation of
+PhysioMotion4D experiment scripts. The test suite is:
 
-- ✅ Comprehensive - Covers all 10 experiment subdirectories (8 active, 2 disabled)
-- ✅ Well-documented - Multiple documentation files and guides
-- ✅ Excluded from CI/CD - Properly marked and filtered
-- ✅ Flexible - Run all experiments or individual ones
-- ✅ Maintainable - Clear structure and extensible design
-- ✅ Production-ready - Error handling, timeouts, reporting
+- Comprehensive - Covers all 10 experiment subdirectories (8 active, 2 disabled)
+- Well-documented - Multiple documentation files and guides
+- Excluded from CI/CD - Properly marked and filtered
+- Flexible - Run all experiments or individual ones
+- Maintainable - Clear structure and extensible design
+- Production-ready - Error handling, timeouts, reporting
 
-The tests serve as both validation tools and living documentation of the experiment workflows.
+The tests serve as both validation tools and living documentation of the
+experiment workflows.
 
 ---
 
-**Last Updated:** 2026-01-25
-**Implementation By:** AI Assistant (Claude Sonnet 4.5)
+**Last Updated:** 2026-05-13
 **Review Status:** Pending human review
 
 ## Change Log
 
 ### 2026-01-25 - Initial Implementation
 - Created comprehensive test suite for 8 experiment subdirectories
-- Implemented automated notebook execution with nbconvert
+- Implemented automated script execution as subprocesses
 - Added extensive documentation and guides
 
 ### 2026-01-25 - Disabled Non-Ready Experiments
-- Disabled `DisplacementField_To_USD` (notebooks not ready)
-- Disabled `Lung-VesselsAirways` (notebooks not ready)
+- Disabled `DisplacementField_To_USD` (scripts not ready)
+- Disabled `Lung-VesselsAirways` (scripts not ready)
 - Tests commented out but can be easily re-enabled
 - Updated documentation to reflect active status
 - 6 experiments remain active (~25 hours total runtime)
@@ -371,15 +377,15 @@ The tests serve as both validation tools and living documentation of the experim
 - Enhanced protection against CI/CD execution
 - Updated all documentation to reflect new flag requirement
 
-### 2026-01-25 - Enforced Sequential Notebook Execution
+### 2026-01-25 - Enforced Sequential Script Execution
 - Added explicit sequential execution enforcement within each experiment test
 - Added `@pytest.mark.xdist_group` markers for pytest-xdist compatibility
-- Modified `run_experiment_notebooks()` to stop on first failure
+- Modified `run_experiment_scripts()` to stop on first failure
 - Added detailed execution order documentation to each test function
 - Added pytest-xdist>=3.0.0 to test dependencies
-- Ensures notebooks run in strict alphanumeric order even with parallel workers
+- Ensures scripts run in strict alphanumeric order even with parallel workers
 - Different experiment subdirectories CAN run in parallel safely
-- Notebooks within a subdirectory CANNOT run in parallel or out of order
+- Scripts within a subdirectory CANNOT run in parallel or out of order
 
 ### 2026-01-30 - Added Heart-Create_Statistical_Model and Renamed Heart-Model_To_Patient
 - Added `Heart-Create_Statistical_Model` experiment (PCA shape model creation with SlicerSALT)
@@ -390,3 +396,10 @@ The tests serve as both validation tools and living documentation of the experim
 - Updated data/KCL-Heart-Model/README.md to reference new experiment
 - Total active experiments: 8 (up from 6)
 - Total estimated runtime: ~32 hours (up from ~25 hours)
+
+### 2026-05-13 - Standardized On Percent-Cell Python Scripts
+- All experiment sources are `.py` files with `# %%` cell markers
+- Test runner uses a plain Python subprocess to execute each script
+- Helper function names: `get_scripts_in_subdir`, `execute_script`,
+  `run_experiment_scripts`
+- Documentation aligned repository-wide
diff --git a/tests/GITHUB_WORKFLOWS.md b/tests/GITHUB_WORKFLOWS.md
index 25d42d0..c2fa0ab 100644
--- a/tests/GITHUB_WORKFLOWS.md
+++ b/tests/GITHUB_WORKFLOWS.md
@@ -251,7 +251,7 @@ These tests run automatically on a schedule via `.github/workflows/test-slow.yml
 
 ### Unit Tests Coverage
 - **Flag**: `unittests`
-- **When**: Ubuntu + Python 3.10 only
+- **When**: Ubuntu + Python 3.11 only
 - **Upload**: After unit tests complete
 
 ### Integration Tests Coverage
@@ -261,7 +261,7 @@ These tests run automatically on a schedule via `.github/workflows/test-slow.yml
 
 ### GPU Tests Coverage
 - **Flag**: `gpu-tests`
-- **When**: Python 3.10 on self-hosted GPU runners
+- **When**: Python 3.11 on self-hosted GPU runners
 - **Upload**: After GPU tests complete
 
 ### Slow Tests Coverage
@@ -285,14 +285,14 @@ Pull Request Created
 ┌───────────────────────────────────────────────┐
 │ Unit Tests (test job)                         │
 │ - Runs on Ubuntu, Windows, macOS              │
-│ - Python 3.10, 3.11, 3.12                    │
+│ - Python 3.11, 3.12                          │
 │ - Fast, no external data                      │
 └───────────────────────────────────────────────┘
     ↓
 ┌───────────────────────────────────────────────┐
 │ Integration Tests (test-with-data job)        │
 │ - Runs on Ubuntu only                         │
-│ - Python 3.10                                 │
+│ - Python 3.11                                 │
 │                                               │
 │ 1. Cache Check (tests/data/, tests/results/) │
 │    ↓                                          │
diff --git a/tests/PARALLEL_EXECUTION_GUIDE.md b/tests/PARALLEL_EXECUTION_GUIDE.md
index 7fb5430..a66ad5a 100644
--- a/tests/PARALLEL_EXECUTION_GUIDE.md
+++ b/tests/PARALLEL_EXECUTION_GUIDE.md
@@ -1,32 +1,35 @@
 # Parallel Execution Guide for Experiment Tests
 
-This guide explains how parallel execution works with experiment tests and how notebook dependencies are enforced.
+This guide explains how parallel execution works with experiment tests and how
+script dependencies are enforced.
 
 ## Overview
 
-Experiment tests support parallel execution at the **subdirectory level** while maintaining **strict sequential execution** of notebooks within each subdirectory.
+Experiment tests support parallel execution at the **subdirectory level** while
+maintaining **strict sequential execution** of scripts within each
+subdirectory.
 
 ## Execution Model
 
 ### Sequential Within Subdirectories (Enforced)
 
-Within each experiment subdirectory, notebooks run in **strict alphanumeric order**:
+Within each experiment subdirectory, scripts run in **strict alphanumeric order**:
 
 ```python
-# Example: Heart-GatedCT_To_USD notebooks
-1. 0-download_and_convert_4d_to_3d.ipynb  ← Must complete first
-2. 1-register_images.ipynb                 ← Uses data from step 1
-3. 2-generate_segmentation.ipynb           ← Uses data from step 2
-4. 3-transform_dynamic_and_static_contours.ipynb  ← Uses data from step 3
-5. 4-merge_dynamic_and_static_usd.ipynb    ← Uses data from step 4
+# Example: Heart-GatedCT_To_USD scripts
+1. 0-download_and_convert_4d_to_3d.py            # Must complete first
+2. 1-register_images.py                          # Uses data from step 1
+3. 2-generate_segmentation.py                    # Uses data from step 2
+4. 3-transform_dynamic_and_static_contours.py    # Uses data from step 3
+5. 4-merge_dynamic_and_static_usd.py             # Uses data from step 4
 # ... and so on
 ```
 
 **Key guarantees:**
-- ✅ Notebooks execute in a Python `for` loop (inherently sequential)
-- ✅ Each notebook must complete before the next begins
-- ✅ **Execution stops on first failure** to prevent cascading errors
-- ✅ This behavior is enforced even when using `pytest -n` (multiple workers)
+- Scripts execute in a Python `for` loop (inherently sequential)
+- Each script must complete before the next begins
+- **Execution stops on first failure** to prevent cascading errors
+- This behavior is enforced even when using `pytest -n` (multiple workers)
 
 ### Parallel Across Subdirectories (Allowed)
 
@@ -50,22 +53,23 @@ Each test function is treated as an atomic unit by pytest-xdist:
 @pytest.mark.xdist_group(name='experiment_heart_gated_ct')
 def test_experiment_heart_gated_ct_to_usd():
     """Test function is atomic - xdist won't split it"""
-    run_experiment_notebooks('Heart-GatedCT_To_USD', ...)
+    run_experiment_scripts('Heart-GatedCT_To_USD', ...)
 ```
 
-pytest-xdist **cannot** and **will not** parallelize the code inside a test function. The entire function runs on a single worker.
+pytest-xdist **cannot** and **will not** parallelize the code inside a test
+function. The entire function runs on a single worker.
 
 ### 2. Sequential Loop Enforcement
 
-Inside `run_experiment_notebooks()`, notebooks run in a standard Python loop:
+Inside `run_experiment_scripts()`, scripts run in a standard Python loop:
 
 ```python
-for i, notebook in enumerate(notebooks, 1):
-    # Execute notebook i
-    result = execute_notebook(notebook, timeout=timeout_per_notebook)
-    
+for i, script in enumerate(scripts, 1):
+    # Execute script i
+    result = execute_script(script, timeout=timeout_per_script)
+
     if not result['success']:
-        # STOP on failure - don't execute remaining notebooks
+        # STOP on failure - don't execute remaining scripts
         break
 ```
 
@@ -85,21 +89,22 @@ def test_experiment_heart_gated_ct_to_usd():
     ...
 ```
 
-This ensures tests in the same group don't run in parallel (though currently each test has a unique group, allowing all to run in parallel).
+This ensures tests in the same group don't run in parallel (though currently
+each test has a unique group, allowing all to run in parallel).
 
 ### 4. Fail-Fast Behavior
 
-When a notebook fails, execution stops immediately:
+When a script fails, execution stops immediately:
 
 ```python
 if not result['success']:
-    print('⚠️ Stopping execution: notebook failed')
-    print('Remaining notebooks in this experiment will not run.')
-    break  # Stop the loop - don't run remaining notebooks
+    print('Stopping execution: script failed')
+    print('Remaining scripts in this experiment will not run.')
+    break  # Stop the loop - don't run remaining scripts
 ```
 
 This prevents:
-- Wasting time on dependent notebooks that will fail
+- Wasting time on dependent scripts that will fail
 - Cascading errors from missing dependencies
 - Confusing error messages from downstream failures
 
@@ -112,9 +117,9 @@ This prevents:
 pytest tests/test_experiments.py -v --run-experiments
 
 # Behavior:
-# - Colormap experiment runs (all notebooks in order)
-# - Then Reconstruct4DCT runs (all notebooks in order)
-# - Then Heart-VTKSeries runs (all notebooks in order)
+# - Colormap experiment runs (all scripts in order)
+# - Then Reconstruct4DCT runs (all scripts in order)
+# - Then Heart-VTKSeries runs (all scripts in order)
 # - ... and so on
 ```
 
@@ -125,8 +130,8 @@ pytest tests/test_experiments.py -v --run-experiments
 pytest tests/test_experiments.py -v -n 2 --run-experiments
 
 # Possible behavior:
-# Worker 1: Colormap experiment (notebooks sequential within)
-# Worker 2: Heart-VTKSeries experiment (notebooks sequential within)
+# Worker 1: Colormap experiment (scripts sequential within)
+# Worker 2: Heart-VTKSeries experiment (scripts sequential within)
 # When Worker 1 finishes, it picks up Reconstruct4DCT
 # When Worker 2 finishes, it picks up Heart-Statistical_Model_To_Patient
 # ... and so on
@@ -141,17 +146,17 @@ pytest tests/test_experiments.py -v -n auto --run-experiments
 # Behavior:
 # - Spawns one worker per CPU core
 # - Distributes test functions (subdirectories) across workers
-# - Within each worker, notebooks run sequentially
+# - Within each worker, scripts run sequentially
 ```
 
 ## Dependency Management
 
 ### Within-Subdirectory Dependencies
 
-Notebooks in the same subdirectory often have dependencies:
+Scripts in the same subdirectory often have dependencies:
 
-```
-0-download.ipynb → 1-process.ipynb → 2-analyze.ipynb → 3-visualize.ipynb
+```text
+0-download.py -> 1-process.py -> 2-analyze.py -> 3-visualize.py
 ```
 
 **Protected by:**
@@ -164,9 +169,9 @@ Notebooks in the same subdirectory often have dependencies:
 Different subdirectories should be independent:
 
 ```
-Colormap-VTK_To_USD/  ← Independent
-Heart-GatedCT_To_USD/ ← Independent
-Lung-GatedCT_To_USD/  ← Independent
+Colormap-VTK_To_USD/    # Independent
+Heart-GatedCT_To_USD/   # Independent
+Lung-GatedCT_To_USD/    # Independent
 ```
 
 **Best practices:**
@@ -208,16 +213,16 @@ Experiments generate large files:
 
 ## Troubleshooting
 
-### Notebooks Running Out of Order
+### Scripts Running Out of Order
 
-**Symptom:** Later notebooks fail due to missing dependencies
+**Symptom:** Later scripts fail due to missing dependencies
 
 **Diagnosis:**
-1. Check that notebooks have numeric prefixes (0-, 1-, 2-)
-2. Verify `sorted()` in `get_notebooks_in_subdir()` is working
+1. Check that scripts have numeric prefixes (0-, 1-, 2-)
+2. Verify `sorted()` in `get_scripts_in_subdir()` is working
 3. Check for parallel execution attempts (shouldn't happen)
 
-**Fix:** Notebooks are sorted alphanumerically. Use proper prefixes.
+**Fix:** Scripts are sorted alphanumerically. Use proper prefixes.
 
 ### Parallel Execution Causing Errors
 
@@ -240,12 +245,12 @@ Experiments generate large files:
 **Diagnosis:**
 1. Check resource limits (memory, GPU)
 2. Look for deadlocks or race conditions
-3. Verify notebooks don't have infinite loops
+3. Verify scripts don't have infinite loops
 
 **Fix:**
 - Reduce worker count (`-n 1` for serial)
 - Increase timeouts if needed
-- Fix problematic notebooks
+- Fix problematic scripts
 
 ## Verification
 
@@ -256,13 +261,13 @@ Experiments generate large files:
 pytest tests/test_experiments.py::test_experiment_heart_gated_ct_to_usd -v -s --run-experiments
 
 # Look for output like:
-# --- Notebook 1/7 ---
-# Sequential execution: notebook 1 must complete before 2 starts
-# ... (notebook 1 output) ...
-# ✅ Successfully executed: 0-download_and_convert_4d_to_3d.ipynb
-# --- Notebook 2/7 ---
-# Sequential execution: notebook 2 must complete before 3 starts
-# ... (notebook 2 output) ...
+# --- Script 1/7 ---
+# Sequential execution: script 1 must complete before 2 starts
+# ... (script 1 output) ...
+# Successfully executed: 0-download_and_convert_4d_to_3d.py
+# --- Script 2/7 ---
+# Sequential execution: script 2 must complete before 3 starts
+# ... (script 2 output) ...
 ```
 
 ### Test Parallel Execution
@@ -279,28 +284,30 @@ pytest tests/test_experiments.py -v -n 2 --run-experiments
 ### Test Fail-Fast Behavior
 
 ```bash
-# Temporarily break a middle notebook
-# Run the test and verify remaining notebooks don't execute
+# Temporarily break a middle script
+# Run the test and verify remaining scripts don't execute
 pytest tests/test_experiments.py::test_experiment_heart_gated_ct_to_usd -v -s --run-experiments
 
 # Should see:
-# ❌ Failed to execute: 2-generate_segmentation.ipynb
-# ⚠️ Stopping execution: 2-generate_segmentation.ipynb failed
-# Remaining notebooks in this experiment will not run.
+# Failed to execute: 2-generate_segmentation.py
+# Stopping execution: 2-generate_segmentation.py failed
+# Remaining scripts in this experiment will not run.
 ```
 
 ## Summary
 
 | Aspect | Behavior | Enforcement |
 |--------|----------|-------------|
-| Notebooks within subdirectory | **Sequential, strict order** | Python `for` loop, fail-fast |
+| Scripts within subdirectory | **Sequential, strict order** | Python `for` loop, fail-fast |
 | Subdirectories (test functions) | **Can run in parallel** | pytest-xdist distribution |
-| Notebook execution | **One at a time per test** | Standard function execution |
+| Script execution | **One at a time per test** | Standard function execution |
 | Failure handling | **Stop on first error** | `break` statement in loop |
 | Worker assignment | **One test per worker** | pytest-xdist scheduling |
-| Order enforcement | **Alphanumeric sorting** | `sorted(glob('*.ipynb'))` |
+| Order enforcement | **Alphanumeric sorting** | `sorted(glob('*.py'))` |
 
-**Bottom line:** Your notebooks within each subdirectory will ALWAYS run in order, even with `pytest -n`. The parallelism only occurs at the subdirectory/test level, not within a test.
+**Bottom line:** Your scripts within each subdirectory will ALWAYS run in
+order, even with `pytest -n`. The parallelism only occurs at the
+subdirectory/test level, not within a test.
 
 ## Related Documentation
 
@@ -310,4 +317,5 @@ pytest tests/test_experiments.py::test_experiment_heart_gated_ct_to_usd -v -s --
 
 ---
 
-**Key Takeaway:** Parallel execution is safe and won't break dependencies. Notebooks in each subdirectory run sequentially no matter what.
+**Key Takeaway:** Parallel execution is safe and won't break dependencies.
+Scripts in each subdirectory run sequentially no matter what.
diff --git a/tests/README.md b/tests/README.md
index 50e95cd..bcb1371 100644
--- a/tests/README.md
+++ b/tests/README.md
@@ -2,16 +2,16 @@
 
 This directory contains comprehensive test suites for the PhysioMotion4D package, validating the complete medical imaging to Omniverse pipeline.
 
-## 📚 Documentation
+## Documentation
 
 - **[TESTING_GUIDE.md](TESTING_GUIDE.md)** - Comprehensive testing guide with setup, troubleshooting, and best practices
 - **[GITHUB_WORKFLOWS.md](GITHUB_WORKFLOWS.md)** - CI/CD documentation and GitHub Actions workflow details
-- **[EXPERIMENT_TESTS_GUIDE.md](EXPERIMENT_TESTS_GUIDE.md)** - Guide for running experiment notebook tests
+- **[EXPERIMENT_TESTS_GUIDE.md](EXPERIMENT_TESTS_GUIDE.md)** - Guide for running experiment script tests
 - **[PARALLEL_EXECUTION_GUIDE.md](PARALLEL_EXECUTION_GUIDE.md)** - How parallel execution works with experiment tests
 - **[EXPERIMENT_FLAG_USAGE.md](EXPERIMENT_FLAG_USAGE.md)** - Details on the --run-experiments flag
 - **[TEST_FIXES_SUMMARY.md](TEST_FIXES_SUMMARY.md)** - Recent bug fixes and known issues
 
-## 🧪 Test Categories
+## Test Categories
 
 ### Data Pipeline Tests
 - **`test_download_heart_data.py`** - Automatic data download with fallback logic
@@ -34,17 +34,17 @@ This directory contains comprehensive test suites for the PhysioMotion4D package
 - **`test_usd_time_preservation.py`** - Time-varying data validation
 
 ### Experiment Tests (EXTREMELY SLOW - Manual Only)
-- **`test_experiments.py`** - End-to-end experiment notebook execution (hours to complete)
-  - 🔒 **Opt-in only** - Requires `--run-experiments` flag to run
-  - ⚠️ **NOT included in CI/CD** - Never runs in automated workflows
-  - ⚠️ **Automatically skipped** - Won't run with `pytest tests/` unless flag is set
-  - Runs all notebooks in `experiments/` subdirectories
+- **`test_experiments.py`** - End-to-end experiment script execution (hours to complete)
+  - **Opt-in only** - Requires `--run-experiments` flag to run
+  - **NOT included in CI/CD** - Never runs in automated workflows
+  - **Automatically skipped** - Won't run with `pytest tests/` unless flag is set
+  - Runs every `*.py` script in each `experiments/` subdirectory
   - Each subdirectory gets its own test
-  - Notebooks run in alphanumeric order
+  - Scripts run in alphanumeric order
   - Requires GPU, CUDA, and all dependencies installed
-  - 📖 **See [EXPERIMENT_TESTS_GUIDE.md](EXPERIMENT_TESTS_GUIDE.md) for detailed usage instructions**
+  - See [EXPERIMENT_TESTS_GUIDE.md](EXPERIMENT_TESTS_GUIDE.md) for detailed usage instructions
 
-## 📂 Directory Structure
+## Directory Structure
 
 ```
 tests/
@@ -56,7 +56,7 @@ tests/
 └── *.py                     # Test modules
 ```
 
-## 🚀 Quick Start
+## Quick Start
 
 ### Git LFS (required for tests)
 
@@ -111,13 +111,13 @@ pytest tests/ --cov=src/physiomotion4d --cov-report=html
 pytest tests/test_usd_merge.py::TestUSDMerge::test_merge_usd_files_copy_method -v
 ```
 
-> 💡 **For detailed instructions**, see [TESTING_GUIDE.md](TESTING_GUIDE.md)
+**For detailed instructions**, see [TESTING_GUIDE.md](TESTING_GUIDE.md)
 
-## ⏱️ Test Timing Reports
+## Test Timing Reports
 
 All test runs automatically generate a comprehensive timing report at the end showing individual test durations, session time, and pass/fail/skip counts. The report separates regular tests from experiment tests and highlights the slowest tests.
 
-## ⚙️ Test Configuration
+## Test Configuration
 
 ### Global Settings
 - **Timeout**: 900 seconds (15 minutes) per test
@@ -150,30 +150,30 @@ test_segment_chest_total_segmentator ────→ test_contour_tools
 
 Fixtures in `conftest.py` automatically manage these dependencies.
 
-## 🔄 Continuous Integration
+## Continuous Integration
 
 Tests automatically run on pull requests via GitHub Actions. The CI workflow:
 
-- ✅ **Runs fast tests** - USD utilities, data conversion, basic validation
-- ❌ **Skips slow tests** - Registration and segmentation (too slow for CI)
-- ❌ **Automatically skips experiment tests** - Protected by `--run-experiments` flag requirement
-- ✅ **Caches test data** - Speeds up subsequent runs
-- ✅ **Generates coverage** - Reports uploaded to Codecov
+- **Runs fast tests** - USD utilities, data conversion, basic validation
+- **Skips slow tests** - Registration and segmentation (too slow for CI)
+- **Automatically skips experiment tests** - Protected by `--run-experiments` flag requirement
+- **Caches test data** - Speeds up subsequent runs
+- **Generates coverage** - Reports uploaded to Codecov
 
 **CI Configuration:**
 - Platforms: Ubuntu, Windows, macOS
-- Python versions: 3.10, 3.11, 3.12
+- Python versions: 3.11, 3.12
 - Target coverage: >70%
 - Protection: Experiment tests require `--run-experiments` flag (never used in CI/CD)
 
-> 📖 **For detailed CI/CD information**, see [GITHUB_WORKFLOWS.md](GITHUB_WORKFLOWS.md)
+**For detailed CI/CD information**, see [GITHUB_WORKFLOWS.md](GITHUB_WORKFLOWS.md)
 
-## 🛠️ Troubleshooting
+## Troubleshooting
 
 ### Common Issues
 
 **Problem: HTTP 404 when downloading data**
-- ✅ **Fixed!** Tests now check `data/Slicer-Heart-CT/` first
+- Tests now check `data/Slicer-Heart-CT/` first
 - Place `TruncalValve_4DCT.seq.nrrd` there to avoid download
 
 **Problem: Out of memory errors**
@@ -186,12 +186,12 @@ Tests automatically run on pull requests via GitHub Actions. The CI workflow:
 - Override with: `@pytest.mark.timeout(1800)` decorator or use `-o timeout=1800`
 
 **Problem: Fixture naming errors**
-- ✅ **Fixed!** Use correct fixture names from `conftest.py`
+- Use correct fixture names from `conftest.py`
 - Don't import from other test files
 
-> 🔍 **For complete troubleshooting guide**, see [TESTING_GUIDE.md](TESTING_GUIDE.md#troubleshooting)
+**For complete troubleshooting guide**, see [TESTING_GUIDE.md](TESTING_GUIDE.md#troubleshooting)
 
-## 📝 Adding New Tests
+## Adding New Tests
 
 When creating new tests:
 
@@ -202,9 +202,9 @@ When creating new tests:
 5. **Organize outputs**: Save results to `tests/results/<module_name>/`
 6. **Update docs**: Add test description to this README
 
-> 💡 **For best practices**, see [TESTING_GUIDE.md](TESTING_GUIDE.md#best-practices)
+**For best practices**, see [TESTING_GUIDE.md](TESTING_GUIDE.md#best-practices)
 
-## 📊 Test Data
+## Test Data
 
 ### Slicer-Heart-CT Dataset
 - **Source**: [Slicer-Heart-CT GitHub](https://github.com/Slicer-Heart-CT/Slicer-Heart-CT)
@@ -219,7 +219,7 @@ Tests automatically:
 3. Download from GitHub if needed
 4. Skip gracefully if unavailable
 
-## 📚 Additional Resources
+## Additional Resources
 
 - **Detailed Testing Guide**: [TESTING_GUIDE.md](TESTING_GUIDE.md)
 - **Experiment Tests Guide**: [EXPERIMENT_TESTS_GUIDE.md](EXPERIMENT_TESTS_GUIDE.md)
diff --git a/tests/TESTING_GUIDE.md b/tests/TESTING_GUIDE.md
index 78f8fa1..3807935 100644
--- a/tests/TESTING_GUIDE.md
+++ b/tests/TESTING_GUIDE.md
@@ -114,10 +114,10 @@ Tests automatically manage data with intelligent fallback:
 4. **Skip test if unavailable**: Graceful failure with helpful message
 
 This approach:
-- ✅ Avoids re-downloading 1.2GB file multiple times
-- ✅ Reuses existing project data
-- ✅ Works offline if data already present
-- ✅ Provides clear error messages
+- Avoids re-downloading 1.2GB file multiple times
+- Reuses existing project data
+- Works offline if data already present
+- Provides clear error messages
 
 ### Slicer-Heart-CT Dataset
 
@@ -154,7 +154,7 @@ tests/
 
 **Problem**: HTTP 404 error when downloading data
 
-**Solution**: ✅ **Fixed!** Tests now automatically:
+**Solution**: **Fixed!** Tests now automatically:
 1. Check for existing data in `data/Slicer-Heart-CT/`
 2. Copy from main data directory if available
 3. Only attempt download if not found locally
@@ -191,7 +191,7 @@ tests/
 
 **Problem**: `TypeError: in method 'itkSize3___getitem__', argument 2 of type 'unsigned long'`
 
-**Solution**: ✅ **Fixed!** ITK Size objects now properly converted to tuples before numpy indexing:
+**Solution**: **Fixed!** ITK Size objects now properly converted to tuples before numpy indexing:
 ```python
 size_itk = itk.size(image)
 size = (int(size_itk[0]), int(size_itk[1]), int(size_itk[2]))
@@ -201,7 +201,7 @@ size = (int(size_itk[0]), int(size_itk[1]), int(size_itk[2]))
 
 **Problem**: `NameError: name 'segmenter' is not defined`
 
-**Solution**: ✅ **Fixed!** Tests now use correct fixture names:
+**Solution**: **Fixed!** Tests now use correct fixture names:
 - `segmenter_total_segmentator` for TotalSegmentator
 - `registrar_ants` for ANTs
 - `registrar_icon` for ICON
@@ -210,7 +210,7 @@ size = (int(size_itk[0]), int(size_itk[1]), int(size_itk[2]))
 
 **Problem**: Tests expecting `DisplacementFieldTransform` but getting `CompositeTransform`
 
-**Solution**: ✅ **Fixed!** Tests now accept both types since registration may return either depending on whether initial transforms are provided.
+**Solution**: **Fixed!** Tests now accept both types since registration may return either depending on whether initial transforms are provided.
 
 ## CI/CD Integration
 
@@ -230,7 +230,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v4
         with:
-          python-version: '3.10'
+          python-version: '3.11'
 
       - name: Install dependencies
         run: |
@@ -284,7 +284,7 @@ def test_something(segmenter_total_segmentator, test_images):
 
 **Bad** - Don't import from other test files:
 ```python
-from test_segment_chest import segmenter  # ❌ Will fail
+from test_segment_chest import segmenter  # Will fail
 ```
 
 ### Memory-Intensive Tests
@@ -361,10 +361,10 @@ rm -rf tests/data/Slicer-Heart-CT/
 
 Tests automatically run on pull requests via `.github/workflows/ci.yml`. The workflow:
 
-- ✅ Runs fast tests (USD, conversion, basic validation)
-- ❌ Skips slow tests (registration, segmentation)
-- ✅ Caches test data to speed up subsequent runs
-- ✅ Generates coverage reports
+- Runs fast tests (USD, conversion, basic validation)
+- Skips slow tests (registration, segmentation)
+- Caches test data to speed up subsequent runs
+- Generates coverage reports
 
 See `tests/GITHUB_WORKFLOWS.md` for detailed CI/CD documentation.
 
diff --git a/tests/baselines/tutorial_01_heart_gated_ct_to_usd/contours_3d.png b/tests/baselines/tutorial_01_heart_gated_ct_to_usd/contours_3d.png
deleted file mode 100644
index 6073c90..0000000
Binary files a/tests/baselines/tutorial_01_heart_gated_ct_to_usd/contours_3d.png and /dev/null differ
diff --git a/tests/baselines/tutorial_01_heart_gated_ct_to_usd/reference_frame_axial.png b/tests/baselines/tutorial_01_heart_gated_ct_to_usd/reference_frame_axial.png
deleted file mode 100644
index 1dc8534..0000000
Binary files a/tests/baselines/tutorial_01_heart_gated_ct_to_usd/reference_frame_axial.png and /dev/null differ
diff --git a/tests/baselines/tutorial_01_heart_gated_ct_to_usd/segmentation_overlay.png b/tests/baselines/tutorial_01_heart_gated_ct_to_usd/segmentation_overlay.png
deleted file mode 100644
index 70a027a..0000000
Binary files a/tests/baselines/tutorial_01_heart_gated_ct_to_usd/segmentation_overlay.png and /dev/null differ
diff --git a/tests/baselines/tutorial_02_ct_to_vtk/segmentation_overlay.png b/tests/baselines/tutorial_02_ct_to_vtk/segmentation_overlay.png
deleted file mode 100644
index a446edd..0000000
Binary files a/tests/baselines/tutorial_02_ct_to_vtk/segmentation_overlay.png and /dev/null differ
diff --git a/tests/baselines/tutorial_02_ct_to_vtk/vtk_surfaces.png b/tests/baselines/tutorial_02_ct_to_vtk/vtk_surfaces.png
deleted file mode 100644
index 26e616d..0000000
Binary files a/tests/baselines/tutorial_02_ct_to_vtk/vtk_surfaces.png and /dev/null differ
diff --git a/tests/baselines/tutorial_05_vtk_to_usd/usd_mesh_rendering.png b/tests/baselines/tutorial_05_vtk_to_usd/usd_mesh_rendering.png
deleted file mode 100644
index 6299c43..0000000
Binary files a/tests/baselines/tutorial_05_vtk_to_usd/usd_mesh_rendering.png and /dev/null differ
diff --git a/tests/conftest.py b/tests/conftest.py
index cc562fb..4449bad 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -7,8 +7,6 @@
 """
 
 import os
-import urllib.error
-import urllib.request
 from datetime import datetime, timedelta
 from pathlib import Path
 from typing import Any, Optional
@@ -18,6 +16,7 @@
 
 from physiomotion4d.contour_tools import ContourTools
 from physiomotion4d.convert_nrrd_4d_to_3d import ConvertNRRD4DTo3D
+from physiomotion4d.data_download_tools import DataDownloadTools
 from physiomotion4d.register_images_ants import RegisterImagesANTs
 from physiomotion4d.register_images_greedy import RegisterImagesGreedy
 from physiomotion4d.register_images_icon import RegisterImagesICON
@@ -297,39 +296,40 @@ def _format_duration(seconds: float) -> str:
 def test_directories() -> dict[str, Path]:
     """Set up test directories for data and results."""
     data_dir = Path(__file__).parent.parent / "data" / "test"
+    slicer_heart_data_dir = data_dir / "slicer_heart"
+    slicer_heart_small_data_dir = data_dir / "slicer_heart_small"
     output_dir = Path(__file__).parent / "results"
     baselines_dir = Path(__file__).parent / "baselines"
 
     # Create directories if they don't exist
     data_dir.mkdir(parents=True, exist_ok=True)
+    slicer_heart_data_dir.mkdir(parents=True, exist_ok=True)
+    slicer_heart_small_data_dir.mkdir(parents=True, exist_ok=True)
     output_dir.mkdir(parents=True, exist_ok=True)
     baselines_dir.mkdir(parents=True, exist_ok=True)
 
-    return {"data": data_dir, "output": output_dir, "baselines": baselines_dir}
+    return {
+        "data": data_dir,
+        "slicer_heart_data": slicer_heart_data_dir,
+        "slicer_heart_small_data": slicer_heart_small_data_dir,
+        "output": output_dir,
+        "baselines": baselines_dir,
+    }
 
 
 @pytest.fixture(scope="session")
 def download_test_data(test_directories: dict[str, Path]) -> Path:
-    """Download TruncalValve 4D CT data."""
-    data_dir = test_directories["data"]
-    input_image_filename = data_dir / "TruncalValve_4DCT.seq.nrrd"
-
-    # Check if file already exists
-    if input_image_filename.exists():
-        print(f"\nData file already exists: {input_image_filename}")
-        return input_image_filename
-
-    # Try to download if not found locally
-    input_image_url = "https://github.com/SlicerHeart/SlicerHeart/releases/download/TestingData/TruncalValve_4DCT.seq.nrrd"
-    print(f"\nDownloading TruncalValve 4D CT data from {input_image_url}...")
+    """Download Slicer-Heart-CT data."""
+    data_dir = test_directories["slicer_heart_data"]
 
     try:
-        urllib.request.urlretrieve(input_image_url, str(input_image_filename))
-        print(f"Downloaded to {input_image_filename}")
-    except urllib.error.URLError as e:
+        input_image_filename = DataDownloadTools.DownloadSlicerHeartCTData(data_dir)
+        print(f"\nSlicer-Heart-CT data ready: {input_image_filename}")
+    except OSError as e:
         msg = (
             f"Could not download test data: {e}. "
-            f"Please manually place TruncalValve_4DCT.seq.nrrd in {data_dir}"
+            "Please manually place "
+            f"{DataDownloadTools.SLICER_HEART_CT_FILENAME} in {data_dir}"
         )
         if os.environ.get("CI"):
             pytest.fail(msg)
@@ -350,7 +350,8 @@ def test_images(
     test_directories: dict[str, Path],
 ) -> list[Any]:
     """Convert and resample 4D NRRD data; return pre-resampled time points."""
-    data_dir = test_directories["data"]
+    data_dir = test_directories["slicer_heart_data"]
+    small_data_dir = test_directories["slicer_heart_small_data"]
 
     # Convert 4D NRRD to 3D time series if not already done
     slice_000 = data_dir / "slice_000.mha"
@@ -359,16 +360,16 @@ def test_images(
         print("\nConverting 4D NRRD to 3D time series...")
         conv = ConvertNRRD4DTo3D()
         conv.load_nrrd_4d(str(download_test_data))
-        conv.save_3d_images(str(data_dir / "slice"))
+        conv.save_3d_images(data_dir, "slice")
     else:
         print("\n3D slice files already exist")
 
-    # Resample each slice_???.mha to 1.5x1.5x1.5 mm and save as slice_???_sml.mha
+    # Resample each slice_???.mha to 1.5x1.5x1.5 mm into slicer_heart_small.
     target_spacing = [1.5, 1.5, 1.5]
     for slice_file in sorted(data_dir.glob("slice_???.mha")):
-        sml_file = slice_file.with_name(slice_file.stem + "_sml.mha")
-        if not sml_file.exists():
-            print(f"\nResampling {slice_file.name} -> {sml_file.name} ...")
+        small_file = small_data_dir / slice_file.name
+        if not small_file.exists():
+            print(f"\nResampling {slice_file.name} -> {small_file.name} ...")
             img = itk.imread(str(slice_file))
             input_spacing = list(img.GetSpacing())
             input_size = list(itk.size(img))
@@ -384,10 +385,10 @@ def test_images(
             resampler.SetOutputOrigin(img.GetOrigin())
             resampler.SetOutputDirection(img.GetDirection())
             resampler.Update()
-            itk.imwrite(resampler.GetOutput(), str(sml_file), compression=True)
+            itk.imwrite(resampler.GetOutput(), str(small_file), compression=True)
     print("\nResampled slice files up to date")
 
-    slice_files = sorted(data_dir.glob("slice_???_sml.mha"))
+    slice_files = sorted(small_data_dir.glob("slice_???.mha"))
     if len(slice_files) < 3:
         pytest.skip("Resampled slice files not found.")
 
@@ -404,14 +405,14 @@ def test_labelmaps(
 ) -> list[dict[str, Any]]:
     """
     Segment each time point with TotalSegmentator and return result dicts.
-    Labelmaps are cached at data_dir / slice_???_sml_labelmap.mha.
+    Labelmaps are cached under slicer_heart_small_data.
     """
-    data_dir = test_directories["data"]
-    slice_files = sorted(data_dir.glob("slice_???_sml.mha"))
+    small_data_dir = test_directories["slicer_heart_small_data"]
+    slice_files = sorted(small_data_dir.glob("slice_???.mha"))
 
     results: list[dict[str, Any]] = []
     for img, slice_file in zip(test_images, slice_files):
-        labelmap_file = data_dir / f"{slice_file.stem}_labelmap.mha"
+        labelmap_file = slice_file.with_name(f"{slice_file.stem}_labelmap.mha")
         if not labelmap_file.exists():
             print(f"\nSegmenting {slice_file.name} ...")
             result = segmenter_total_segmentator.segment(
@@ -446,13 +447,12 @@ def test_transforms(
     """
     Perform ANTs registration and return results.
     Generates them if not already present, otherwise loads from disk.
-    Transforms are cached in data_dir alongside the slice files.
+    Transforms are cached under slicer_heart_small_data.
     """
-    data_dir = test_directories["data"]
-
+    small_data_dir = test_directories["slicer_heart_small_data"]
     frame_tag = "001_to_007"
-    inverse_transform_path = data_dir / f"ants_inverse_transform_{frame_tag}.hdf"
-    forward_transform_path = data_dir / f"ants_forward_transform_{frame_tag}.hdf"
+    inverse_transform_path = small_data_dir / f"ants_inverse_transform_{frame_tag}.hdf"
+    forward_transform_path = small_data_dir / f"ants_forward_transform_{frame_tag}.hdf"
 
     if inverse_transform_path.exists() and forward_transform_path.exists():
         print("\nLoading existing ANTs registration results...")
diff --git a/tests/test_anatomy_taxonomy.py b/tests/test_anatomy_taxonomy.py
new file mode 100644
index 0000000..2483333
--- /dev/null
+++ b/tests/test_anatomy_taxonomy.py
@@ -0,0 +1,130 @@
+"""Unit tests for :class:`physiomotion4d.AnatomyTaxonomy`.
+
+These tests exercise the pure-data taxonomy in isolation — no ITK, no pxr,
+no GPU. They are fast and run unconditionally in the default test suite.
+"""
+
+from __future__ import annotations
+
+import pytest
+
+from physiomotion4d import AnatomyGroup, AnatomyTaxonomy
+
+
+def test_add_organ_creates_group_lazily() -> None:
+    tax = AnatomyTaxonomy()
+    assert tax.group_names() == []
+
+    tax.add_organ("heart", 51, "heart")
+    assert tax.group_names() == ["heart"]
+    assert tax.labels_in_group("heart") == {51: "heart"}
+
+
+def test_group_names_preserve_insertion_order() -> None:
+    tax = AnatomyTaxonomy()
+    tax.add_organ("lung", 10, "lung_upper_lobe_left")
+    tax.add_organ("heart", 51, "heart")
+    tax.add_organ("bone", 91, "skull")
+    assert tax.group_names() == ["lung", "heart", "bone"]
+
+
+def test_labels_in_group_unknown_returns_empty_dict() -> None:
+    tax = AnatomyTaxonomy()
+    tax.add_organ("heart", 51, "heart")
+    assert tax.labels_in_group("does_not_exist") == {}
+
+
+def test_labels_in_group_returns_copy_not_alias() -> None:
+    """Caller mutation of the returned dict must not corrupt the taxonomy."""
+    tax = AnatomyTaxonomy()
+    tax.add_organ("heart", 51, "heart")
+    snapshot = tax.labels_in_group("heart")
+    snapshot[999] = "bogus"
+    assert tax.labels_in_group("heart") == {51: "heart"}
+
+
+def test_all_labels_merges_every_group() -> None:
+    tax = AnatomyTaxonomy()
+    tax.add_organ("heart", 51, "heart")
+    tax.add_organ("heart", 61, "atrial_appendage_left")
+    tax.add_organ("lung", 10, "lung_upper_lobe_left")
+    assert tax.all_labels() == {
+        51: "heart",
+        61: "atrial_appendage_left",
+        10: "lung_upper_lobe_left",
+    }
+
+
+def test_group_for_label_finds_by_organ_name() -> None:
+    tax = AnatomyTaxonomy()
+    tax.add_organ("heart", 51, "heart")
+    tax.add_organ("heart", 61, "atrial_appendage_left")
+    tax.add_organ("lung", 10, "lung_upper_lobe_left")
+    assert tax.group_for_label("atrial_appendage_left") == "heart"
+    assert tax.group_for_label("lung_upper_lobe_left") == "lung"
+
+
+def test_group_for_label_unknown_falls_back_to_other() -> None:
+    tax = AnatomyTaxonomy()
+    tax.add_organ("heart", 51, "heart")
+    assert tax.group_for_label("not_in_any_group") == "other"
+
+
+def test_group_for_id_finds_by_id() -> None:
+    tax = AnatomyTaxonomy()
+    tax.add_organ("heart", 51, "heart")
+    tax.add_organ("lung", 10, "lung_upper_lobe_left")
+    assert tax.group_for_id(51) == "heart"
+    assert tax.group_for_id(10) == "lung"
+    assert tax.group_for_id(9999) == "other"
+
+
+def test_fill_other_group_only_claims_unassigned_ids() -> None:
+    tax = AnatomyTaxonomy()
+    tax.add_organ("heart", 51, "heart")
+    tax.add_organ("lung", 10, "lung_upper_lobe_left")
+    tax.fill_other_group(id_range=range(1, 12))
+
+    # Existing assignments untouched.
+    assert tax.group_for_id(51) == "heart"
+    assert tax.group_for_id(10) == "lung"
+
+    # Unclaimed ids in [1, 12) all landed under 'other' with synthetic names.
+    other_labels = tax.labels_in_group("other")
+    expected_ids = {1, 2, 3, 4, 5, 6, 7, 8, 9, 11}
+    assert set(other_labels.keys()) == expected_ids
+    for label_id in expected_ids:
+        assert other_labels[label_id] == f"other_{label_id}"
+
+
+def test_fill_other_group_idempotent_for_already_claimed_other() -> None:
+    """Calling fill_other_group twice must not duplicate or overwrite."""
+    tax = AnatomyTaxonomy()
+    tax.add_organ("heart", 51, "heart")
+    tax.fill_other_group(id_range=range(50, 53))
+    snapshot = dict(tax.labels_in_group("other"))
+    tax.fill_other_group(id_range=range(50, 53))
+    assert tax.labels_in_group("other") == snapshot
+
+
+def test_anatomy_group_dataclass_default_organs() -> None:
+    group = AnatomyGroup(name="heart")
+    assert group.name == "heart"
+    assert group.organs == {}
+
+
+def test_segment_anatomy_base_default_taxonomy_seeded() -> None:
+    """SegmentAnatomyBase seeds contrast (135) and soft_tissue (133)."""
+    # Import lazily to avoid pulling itk in test collection if it's unused
+    # by sibling tests in this module.
+    from physiomotion4d import SegmentAnatomyBase
+
+    seg = SegmentAnatomyBase()
+    assert seg.taxonomy.group_for_id(135) == "contrast"
+    assert seg.taxonomy.group_for_id(133) == "soft_tissue"
+    assert seg.label_to_type("contrast") == "contrast"
+    assert seg.label_to_type("soft_tissue") == "soft_tissue"
+
+
+if __name__ == "__main__":  # pragma: no cover
+    pytest.main([__file__, "-v"])
diff --git a/tests/test_convert_nrrd_4d_to_3d.py b/tests/test_convert_nrrd_4d_to_3d.py
index 3ad47cb..d87ed14 100644
--- a/tests/test_convert_nrrd_4d_to_3d.py
+++ b/tests/test_convert_nrrd_4d_to_3d.py
@@ -22,7 +22,7 @@ def test_convert_4d_to_3d(
         download_test_data: Path,
         test_directories: dict[str, Path],
     ) -> None:
-        """Test conversion of 4D NRRD to 3D time series (replicates notebook cell 3)."""
+        """Test conversion of 4D NRRD to 3D time series."""
         output_dir = test_directories["output"] / "convert_nrrd_4d_to_3d"
         output_dir.mkdir(parents=True, exist_ok=True)
 
@@ -32,7 +32,7 @@ def test_convert_4d_to_3d(
         print("\nConverting 4D NRRD to 3D time series...")
         conv = ConvertNRRD4DTo3D()
         conv.load_nrrd_4d(str(input_4d_file))
-        conv.save_3d_images(str(output_dir / "slice"))
+        conv.save_3d_images(output_dir, "slice")
 
         # Verify that slice files were created
         slice_007 = output_dir / "slice_007.mha"
@@ -94,8 +94,7 @@ def test_save_3d_images(
         num_time_points = conv.get_number_of_3d_images()
 
         # Save to a test subdirectory
-        test_output_prefix = output_dir / "test_slice"
-        conv.save_3d_images(str(test_output_prefix))
+        conv.save_3d_images(output_dir, "test_slice")
 
         # Verify files were created
         test_slice_files = list(output_dir.glob("test_slice_*.mha"))
diff --git a/tests/test_convert_vtk_to_usd.py b/tests/test_convert_vtk_to_usd.py
index 9d9735c..d5989a8 100644
--- a/tests/test_convert_vtk_to_usd.py
+++ b/tests/test_convert_vtk_to_usd.py
@@ -484,7 +484,8 @@ def test_static_merge_prim_names_use_data_basename(self, tmp_path: Path) -> None
     # ------------------------------------------------------------------
 
     def test_mask_ids_basic_produces_per_label_prims(self, tmp_path: Path) -> None:
-        """mask_ids must produce one USD prim per label; no unified /Mesh prim."""
+        """mask_ids must produce one USD prim per label grouped under /Anatomy
+        when no segmenter is supplied; no unified /Mesh prim."""
         label_ids = [1, 1, 1, 1, 1, 2, 2, 2, 2]
         mesh = _make_poly(label_ids=label_ids)
         converter = ConvertVTKToUSD(
@@ -494,8 +495,8 @@ def test_mask_ids_basic_produces_per_label_prims(self, tmp_path: Path) -> None:
         )
         stage = converter.convert(str(tmp_path / "out.usd"))
 
-        ventricle = stage.GetPrimAtPath("/World/Heart/ventricle")
-        atrium = stage.GetPrimAtPath("/World/Heart/atrium")
+        ventricle = stage.GetPrimAtPath("/World/Heart/Anatomy/ventricle")
+        atrium = stage.GetPrimAtPath("/World/Heart/Anatomy/atrium")
         assert ventricle.IsValid(), "ventricle prim not found"
         assert atrium.IsValid(), "atrium prim not found"
         assert ventricle.IsA(UsdGeom.Mesh)
@@ -504,6 +505,9 @@ def test_mask_ids_basic_produces_per_label_prims(self, tmp_path: Path) -> None:
         assert list(UsdGeom.Mesh(atrium).GetPointsAttr().GetTimeSamples()) == [0.0]
         # Unified prim must NOT exist when mask_ids is active
         assert not stage.GetPrimAtPath("/World/Heart/Mesh").IsValid()
+        # Materials live under the same /Anatomy group as the meshes
+        assert stage.GetPrimAtPath("/World/Looks/Anatomy/ventricle_material").IsValid()
+        assert stage.GetPrimAtPath("/World/Looks/Anatomy/atrium_material").IsValid()
 
     def test_structured_grid_extracts_surface(self, tmp_path: Path) -> None:
         """StructuredGrid input is surface-extracted when convert_to_surface is true."""
@@ -543,8 +547,8 @@ def test_mask_ids_missing_label_filters_time_codes(self, tmp_path: Path) -> None
         converter._time_codes = [0.0, 1.0, 2.0]
         stage = converter.convert(str(tmp_path / "out.usd"))
 
-        ventricle = stage.GetPrimAtPath("/World/Heart/ventricle")
-        atrium = stage.GetPrimAtPath("/World/Heart/atrium")
+        ventricle = stage.GetPrimAtPath("/World/Heart/Anatomy/ventricle")
+        atrium = stage.GetPrimAtPath("/World/Heart/Anatomy/atrium")
         assert ventricle.IsValid()
         assert atrium.IsValid()
 
@@ -566,7 +570,42 @@ def test_mask_ids_missing_boundary_labels_falls_back(self, tmp_path: Path) -> No
         )
         stage = converter.convert(str(tmp_path / "out.usd"))
 
-        assert stage.GetPrimAtPath("/World/FB/default").IsValid(), (
+        assert stage.GetPrimAtPath("/World/FB/Anatomy/default").IsValid(), (
             "'default' fallback prim missing"
         )
-        assert not stage.GetPrimAtPath("/World/FB/ventricle").IsValid()
+        assert not stage.GetPrimAtPath("/World/FB/Anatomy/ventricle").IsValid()
+
+    def test_mask_ids_groups_by_segmenter_type(self, tmp_path: Path) -> None:
+        """When a segmenter is supplied, labels are grouped under their
+        anatomy type (heart/lung/...) for both meshes and materials."""
+        from physiomotion4d.segment_chest_total_segmentator import (
+            SegmentChestTotalSegmentator,
+        )
+
+        # Label 51 is "heart" (heart group), 10 is "lung_upper_lobe_left" (lung group).
+        label_ids = [51, 51, 51, 51, 51, 10, 10, 10, 10]
+        mesh = _make_poly(label_ids=label_ids)
+        seg = SegmentChestTotalSegmentator()
+        converter = ConvertVTKToUSD(
+            data_basename="Chest",
+            input_polydata=[mesh],
+            mask_ids={51: "heart", 10: "lung_upper_lobe_left"},
+            segmenter=seg,
+        )
+        stage = converter.convert(str(tmp_path / "out.usd"))
+
+        # Meshes grouped under per-type Xforms
+        heart_mesh = stage.GetPrimAtPath("/World/Chest/heart/heart")
+        lung_mesh = stage.GetPrimAtPath("/World/Chest/lung/lung_upper_lobe_left")
+        assert heart_mesh.IsValid(), "heart mesh should land under /heart group"
+        assert lung_mesh.IsValid(), "lung mesh should land under /lung group"
+        assert heart_mesh.IsA(UsdGeom.Mesh)
+        assert lung_mesh.IsA(UsdGeom.Mesh)
+
+        # Materials mirror the mesh hierarchy
+        assert stage.GetPrimAtPath("/World/Looks/heart/heart_material").IsValid()
+        assert stage.GetPrimAtPath(
+            "/World/Looks/lung/lung_upper_lobe_left_material"
+        ).IsValid()
+        # Flat fallback group must NOT be defined when segmenter classifies all labels
+        assert not stage.GetPrimAtPath("/World/Chest/Anatomy").IsValid()
diff --git a/tests/test_download_heart_data.py b/tests/test_download_heart_data.py
index 421ce1e..8a7f534 100644
--- a/tests/test_download_heart_data.py
+++ b/tests/test_download_heart_data.py
@@ -10,6 +10,55 @@
 
 import pytest
 
+from physiomotion4d.data_download_tools import DataDownloadTools
+
+
+class TestDataDownloadTools:
+    """Synthetic tests for dataset verification helpers."""
+
+    def test_verify_slicer_heart_ct_data(self, tmp_path: Path) -> None:
+        """Verify Slicer data by expected `TruncalValve_4DCT.seq.nrrd` filename."""
+        data_file = tmp_path / DataDownloadTools.SLICER_HEART_CT_FILENAME
+
+        assert not DataDownloadTools.VerifySlicerHeartCTData(tmp_path)
+
+        data_file.write_bytes(b"nrrd")
+
+        assert DataDownloadTools.VerifySlicerHeartCTData(tmp_path)
+
+    def test_verify_kcl_heart_model_data(self, tmp_path: Path) -> None:
+        """Verify KCL data by expected average mesh and input mesh filenames."""
+        (tmp_path / "average_mesh.vtk").write_text("# vtk\n")
+        input_meshes = tmp_path / "input_meshes"
+        input_meshes.mkdir()
+
+        assert not DataDownloadTools.VerifyKCLHeartModelData(tmp_path)
+
+        (input_meshes / "01.vtk").write_text("# vtk\n")
+
+        assert DataDownloadTools.VerifyKCLHeartModelData(tmp_path)
+
+    def test_verify_dirlab_4dct_data(self, tmp_path: Path) -> None:
+        """Verify DirLab data by supported Case1 phase image layouts."""
+        case1_dir = tmp_path / "Case1"
+        case1_dir.mkdir()
+
+        assert not DataDownloadTools.VerifyDirLab4DCTData(tmp_path)
+
+        (case1_dir / "case1_T00.mhd").write_text("ObjectType = Image\n")
+
+        assert DataDownloadTools.VerifyDirLab4DCTData(tmp_path)
+
+    def test_verify_chop_valve_4d_data(self, tmp_path: Path) -> None:
+        """Verify CHOP data by expected CT or valve time-series paths."""
+        assert not DataDownloadTools.VerifyCHOPValve4DData(tmp_path)
+
+        ct_dir = tmp_path / "CT"
+        ct_dir.mkdir()
+        (ct_dir / "RVOT28-Dias.nii.gz").write_bytes(b"nii")
+
+        assert DataDownloadTools.VerifyCHOPValve4DData(tmp_path)
+
 
 @pytest.mark.requires_data
 class TestDownloadHeartData:
@@ -18,11 +67,25 @@ class TestDownloadHeartData:
     def test_directories_created(self, test_directories: dict[str, Path]) -> None:
         """Test that directories are created successfully."""
         data_dir = test_directories["data"]
+        slicer_heart_dir = test_directories["slicer_heart_data"]
+        slicer_heart_small_dir = test_directories["slicer_heart_small_data"]
         output_dir = test_directories["output"]
 
         assert data_dir.exists(), f"Data directory not created: {data_dir}"
+        assert slicer_heart_dir.exists(), (
+            f"Slicer-Heart directory not created: {slicer_heart_dir}"
+        )
+        assert slicer_heart_small_dir.exists(), (
+            f"Small Slicer-Heart directory not created: {slicer_heart_small_dir}"
+        )
         assert output_dir.exists(), f"Output directory not created: {output_dir}"
         assert data_dir.is_dir(), f"Data path is not a directory: {data_dir}"
+        assert slicer_heart_dir.is_dir(), (
+            f"Slicer-Heart path is not a directory: {slicer_heart_dir}"
+        )
+        assert slicer_heart_small_dir.is_dir(), (
+            f"Small Slicer-Heart path is not a directory: {slicer_heart_small_dir}"
+        )
         assert output_dir.is_dir(), f"Output path is not a directory: {output_dir}"
 
     def test_data_downloaded(
@@ -33,6 +96,9 @@ def test_data_downloaded(
         """Test that the TruncalValve 4D CT data file is downloaded."""
         data_file = download_test_data
 
+        assert DataDownloadTools.VerifySlicerHeartCTData(
+            test_directories["slicer_heart_data"]
+        )
         assert data_file.exists(), f"Data file not found: {data_file}"
         assert data_file.is_file(), f"Data path is not a file: {data_file}"
 
diff --git a/tests/test_register_time_series_images.py b/tests/test_register_time_series_images.py
index 7f70bca..a8ca5df 100644
--- a/tests/test_register_time_series_images.py
+++ b/tests/test_register_time_series_images.py
@@ -165,8 +165,8 @@ def test_register_time_series_basic(
 
         test_tools = TestTools(
             class_name=self._class_name,
-            results_dir=test_directories["output"],
-            baselines_dir=test_directories["baselines"],
+            results_dir=test_directories["output"] / self._class_name,
+            baselines_dir=test_directories["baselines"] / self._class_name,
         )
 
         test_tools.write_result_transform(
@@ -226,8 +226,8 @@ def test_register_time_series_with_prior(
 
         test_tools = TestTools(
             class_name=self._class_name,
-            results_dir=test_directories["output"],
-            baselines_dir=test_directories["baselines"],
+            results_dir=test_directories["output"] / self._class_name,
+            baselines_dir=test_directories["baselines"] / self._class_name,
         )
 
         test_tools.write_result_transform(
@@ -400,8 +400,8 @@ def test_transform_application_time_series(
         # Save registered image
         test_tools = TestTools(
             class_name=self._class_name,
-            results_dir=test_directories["output"],
-            baselines_dir=test_directories["baselines"],
+            results_dir=test_directories["output"] / self._class_name,
+            baselines_dir=test_directories["baselines"] / self._class_name,
         )
 
         test_tools.write_result_image(
diff --git a/tests/test_segment_chest_total_segmentator.py b/tests/test_segment_chest_total_segmentator.py
index 1fb5ae5..765c2fc 100644
--- a/tests/test_segment_chest_total_segmentator.py
+++ b/tests/test_segment_chest_total_segmentator.py
@@ -30,32 +30,26 @@ def test_segmenter_initialization(
             "Target spacing not set correctly"
         )
 
-        # Check that anatomical structure ID mappings are defined
-        assert len(segmenter_total_segmentator.heart_mask_ids) > 0, (
-            "Heart mask IDs not defined"
-        )
-        assert len(segmenter_total_segmentator.major_vessels_mask_ids) > 0, (
+        # Check that anatomical structure ID mappings are defined via the
+        # shared taxonomy.
+        taxonomy = segmenter_total_segmentator.taxonomy
+        assert len(taxonomy.labels_in_group("heart")) > 0, "Heart mask IDs not defined"
+        assert len(taxonomy.labels_in_group("major_vessels")) > 0, (
             "Major vessels mask IDs not defined"
         )
-        assert len(segmenter_total_segmentator.lung_mask_ids) > 0, (
-            "Lung mask IDs not defined"
-        )
-        assert len(segmenter_total_segmentator.bone_mask_ids) > 0, (
-            "Bone mask IDs not defined"
-        )
-        assert len(segmenter_total_segmentator.soft_tissue_mask_ids) > 0, (
+        assert len(taxonomy.labels_in_group("lung")) > 0, "Lung mask IDs not defined"
+        assert len(taxonomy.labels_in_group("bone")) > 0, "Bone mask IDs not defined"
+        assert len(taxonomy.labels_in_group("soft_tissue")) > 0, (
             "Soft tissue mask IDs not defined"
         )
 
         print("\nSegmenter initialized with correct parameters")
-        print(f"  Heart structures: {len(segmenter_total_segmentator.heart_mask_ids)}")
-        print(
-            f"  Major vessels: {len(segmenter_total_segmentator.major_vessels_mask_ids)}"
-        )
-        print(f"  Lung structures: {len(segmenter_total_segmentator.lung_mask_ids)}")
-        print(f"  Bone structures: {len(segmenter_total_segmentator.bone_mask_ids)}")
+        print(f"  Heart structures: {len(taxonomy.labels_in_group('heart'))}")
+        print(f"  Major vessels: {len(taxonomy.labels_in_group('major_vessels'))}")
+        print(f"  Lung structures: {len(taxonomy.labels_in_group('lung'))}")
+        print(f"  Bone structures: {len(taxonomy.labels_in_group('bone'))}")
         print(
-            f"  Soft tissue structures: {len(segmenter_total_segmentator.soft_tissue_mask_ids)}"
+            f"  Soft tissue structures: {len(taxonomy.labels_in_group('soft_tissue'))}"
         )
 
     def test_segment_single_image(
diff --git a/tests/test_segment_heart_simpleware.py b/tests/test_segment_heart_simpleware.py
index 8d24188..b28abea 100644
--- a/tests/test_segment_heart_simpleware.py
+++ b/tests/test_segment_heart_simpleware.py
@@ -42,12 +42,19 @@ def test_segmenter_initialization(
             "Target spacing should be 1.0 mm for Simpleware"
         )
 
-        assert len(seg.heart_mask_ids) > 0, "Heart mask IDs not defined"
-        assert len(seg.major_vessels_mask_ids) > 0, "Major vessels mask IDs not defined"
-        # ASCardio does not segment lung, bone, soft_tissue (empty dicts)
-        assert seg.lung_mask_ids == {}, "ASCardio does not segment lungs"
-        assert seg.bone_mask_ids == {}, "ASCardio does not segment bone"
-        assert seg.soft_tissue_mask_ids == {}, "ASCardio does not segment soft tissue"
+        taxonomy = seg.taxonomy
+        assert len(taxonomy.labels_in_group("heart")) > 0, "Heart mask IDs not defined"
+        assert len(taxonomy.labels_in_group("major_vessels")) > 0, (
+            "Major vessels mask IDs not defined"
+        )
+        # ASCardio does not segment lung or bone — those groups are never
+        # registered, so labels_in_group returns an empty dict for them.
+        # soft_tissue still contains the base-class placeholder (id 133).
+        assert taxonomy.labels_in_group("lung") == {}, "ASCardio does not segment lungs"
+        assert taxonomy.labels_in_group("bone") == {}, "ASCardio does not segment bone"
+        assert taxonomy.labels_in_group("soft_tissue") == {133: "soft_tissue"}, (
+            "Only the base-class soft_tissue placeholder should be present"
+        )
 
         assert seg.simpleware_exe_path is not None, "Simpleware executable path not set"
         assert seg.simpleware_script_path is not None, "Simpleware script path not set"
@@ -55,8 +62,8 @@ def test_segmenter_initialization(
 
         print("\nSegmenter initialized with correct parameters")
         print(f"  Target spacing: {seg.target_spacing} mm")
-        print(f"  Heart structures: {len(seg.heart_mask_ids)}")
-        print(f"  Major vessels: {len(seg.major_vessels_mask_ids)}")
+        print(f"  Heart structures: {len(taxonomy.labels_in_group('heart'))}")
+        print(f"  Major vessels: {len(taxonomy.labels_in_group('major_vessels'))}")
 
     def test_set_simpleware_executable_path(
         self, segmenter_simpleware: SegmentHeartSimpleware
@@ -93,19 +100,28 @@ def test_segment_single_image(
         result = segmenter_simpleware.segment(input_image, contrast_enhanced_study=True)
 
         assert isinstance(result, dict), "Result should be a dictionary"
+        # The Simpleware segmenter only registers the groups it actually
+        # populates: heart + major_vessels (subclass) and soft_tissue +
+        # contrast (inherited base-class placeholders). lung and bone are
+        # NOT in the result because ASCardio does not segment them; callers
+        # that need those groups must check membership first.
         expected_keys = [
             "labelmap",
-            "lung",
             "heart",
             "major_vessels",
-            "bone",
             "soft_tissue",
-            "other",
             "contrast",
+            "other",
         ]
         for key in expected_keys:
             assert key in result, f"Missing key '{key}' in result"
             assert result[key] is not None, f"Result['{key}'] is None"
+        assert "lung" not in result, (
+            "ASCardio does not segment lung; key must be absent"
+        )
+        assert "bone" not in result, (
+            "ASCardio does not segment bone; key must be absent"
+        )
 
         labelmap = result["labelmap"]
         assert itk.size(labelmap) == itk.size(input_image), "Labelmap size mismatch"
@@ -138,15 +154,15 @@ def test_anatomy_group_masks(
         input_image = test_images[3]
         result = segmenter_simpleware.segment(input_image, contrast_enhanced_study=True)
 
+        # Only assert on groups Simpleware/ASCardio actually populates.
         anatomy_groups = [
-            "lung",
             "heart",
             "major_vessels",
-            "bone",
             "soft_tissue",
             "other",
         ]
         for group in anatomy_groups:
+            assert group in result, f"{group} mask should be present"
             mask = result[group]
             assert mask is not None, f"{group} mask is None"
             mask_arr = itk.array_from_image(mask)
diff --git a/tests/test_tutorials.py b/tests/test_tutorials.py
index 9fcb34e..35990c7 100644
--- a/tests/test_tutorials.py
+++ b/tests/test_tutorials.py
@@ -6,7 +6,7 @@
 
 Screenshot comparison uses the existing ITK-based baseline infrastructure:
 
-1. The tutorial's ``run_tutorial()`` saves PNGs directly to its ``output_dir``.
+1. Each tutorial script saves PNGs directly to its ``OUTPUT_DIR``.
 2. ``TestTools.compare_result_to_baseline_image`` reads each PNG from that
    directory and compares it against a stored baseline with loose tolerances.
 
@@ -21,12 +21,10 @@
 
 from __future__ import annotations
 
+import runpy
 from pathlib import Path
-from types import SimpleNamespace
-from typing import Any, Optional
+from typing import Any
 
-import itk
-import numpy as np
 import pytest
 
 from physiomotion4d.test_tools import TestTools
@@ -36,24 +34,13 @@
 _PX_TOL = 10.0  # per-pixel absolute error (0-255 range)
 _MAX_PX = 2000  # maximum number of pixels allowed above _PX_TOL
 _TOT_TOL = float("inf")  # use the pixel-count criterion only
+_REPO_ROOT = Path(__file__).parent.parent
 
 
-class _FakeMesh:
-    """Minimal mesh double for tutorial screenshot selection tests."""
-
-    def __init__(self, n_points: int) -> None:
-        self.n_points = n_points
-
-
-class _FakeSurfaceExtractable:
-    """Minimal mesh double that records PyVista extraction options."""
-
-    def __init__(self) -> None:
-        self.algorithm: Optional[str] = None
-
-    def extract_surface(self, *, algorithm: str) -> "_FakeSurfaceExtractable":
-        self.algorithm = algorithm
-        return self
+@pytest.fixture(autouse=True)
+def _enable_tutorial_test_mode(monkeypatch: pytest.MonkeyPatch) -> None:
+    """Run tutorials against repo data/test through TestTools mode switching."""
+    monkeypatch.setenv("PHYSIOMOTION_RUNNING_AS_TEST", "1")
 
 
 def _compare_screenshots(
@@ -62,7 +49,7 @@ def _compare_screenshots(
 ) -> None:
     """Read each PNG as itk.Image and compare against baseline."""
     if not screenshots:
-        pytest.fail("No screenshots produced by run_tutorial")
+        pytest.fail("No screenshots produced by tutorial script")
 
     for png_path in screenshots:
         if not png_path.exists():
@@ -75,99 +62,15 @@ def _compare_screenshots(
         ), f"Screenshot baseline mismatch: {png_path.name}"
 
 
-def test_testtools_results_output_dir_override(tmp_path: Path) -> None:
-    """Store result artifacts in an explicit directory when requested."""
-    results_root = tmp_path / "results"
-    output_dir = tmp_path / "tutorial_output"
-    baselines_root = tmp_path / "baselines"
-    tt = TestTools(
-        results_dir=results_root,
-        baselines_dir=baselines_root,
-        class_name="tutorial_example",
-        results_output_dir=output_dir,
-    )
-
-    image = itk.image_from_array(np.zeros((2, 2, 2), dtype=np.uint8))
-    tt.write_result_image(image, "current.mha")
-
-    assert (output_dir / "current.mha").exists()
-    assert not (results_root / "tutorial_example" / "current.mha").exists()
-    assert (baselines_root / "tutorial_example").exists()
-
-
-# -----------------------------------------------------------------------------
-# Tutorial 1 - Heart-Gated CT to Animated USD
-# -----------------------------------------------------------------------------
-
-
-def test_tutorial_01_contour_png_mesh_uses_current_run_results() -> None:
-    """Select current in-memory contours instead of disk VTP outputs."""
-    from tutorials.tutorial_01_heart_gated_ct_to_usd import (
-        _first_current_contour_mesh,
-    )
-
-    transformed_mesh = _FakeMesh(4)
-    reference_mesh = _FakeMesh(8)
-    workflow: Any = SimpleNamespace(
-        _transformed_contours={"all": [transformed_mesh]},
-        _reference_contours={"all": reference_mesh},
-    )
-
-    assert _first_current_contour_mesh(workflow) is transformed_mesh
-
-
-def test_tutorial_01_reference_png_uses_workflow_fixed_image(
-    tmp_path: Path,
-) -> None:
-    """Select the actual workflow reference image over cached slice images."""
-    from tutorials.tutorial_01_heart_gated_ct_to_usd import (
-        _current_reference_image,
-    )
-
-    fixed_image = object()
-    workflow: Any = SimpleNamespace(_fixed_image=fixed_image)
-    (tmp_path / "slice_000.mha").touch()
-
-    assert _current_reference_image(workflow, tmp_path) is fixed_image
-
-
-def test_tutorial_01_overlay_uses_workflow_fixed_segmentation(
-    tmp_path: Path,
-) -> None:
-    """Select the current fixed labelmap over cached slice labelmaps."""
-    from tutorials.tutorial_01_heart_gated_ct_to_usd import (
-        _current_reference_segmentation,
-    )
-
-    labelmap = object()
-    workflow: Any = SimpleNamespace(_fixed_segmentation={"labelmap": labelmap})
-    (tmp_path / "slice_000_labelmap.mha").touch()
-
-    assert _current_reference_segmentation(workflow, tmp_path) is labelmap
-
-
-def test_tutorial_01_overlay_falls_back_to_fixed_image_mask(
-    tmp_path: Path,
-) -> None:
-    """Read fixed_image_mask.mha before stale slice labelmap files."""
-    from tutorials.tutorial_01_heart_gated_ct_to_usd import (
-        _current_reference_segmentation,
-    )
-
-    arr = np.array(
-        [[[1, 0], [0, 1]], [[0, 1], [1, 0]]],
-        dtype=np.uint8,
+def _run_tutorial_script(script_name: str) -> dict[str, Any]:
+    """Run a tutorial script with no command-line arguments."""
+    namespace = runpy.run_path(
+        str(_REPO_ROOT / "tutorials" / script_name),
+        run_name="__main__",
     )
-    mask = itk.image_from_array(arr)
-    fixed_mask_path = tmp_path / "fixed_image_mask.mha"
-    itk.imwrite(mask, str(fixed_mask_path), compression=True)
-    (tmp_path / "slice_000_labelmap.mha").touch()
-
-    workflow: Any = SimpleNamespace()
-    selected = _current_reference_segmentation(workflow, tmp_path)
-
-    assert selected is not None
-    assert tuple(selected.GetLargestPossibleRegion().GetSize()) == (2, 2, 2)
+    results = namespace.get("tutorial_results")
+    assert isinstance(results, dict), f"{script_name} did not set tutorial_results"
+    return results
 
 
 @pytest.mark.tutorial
@@ -176,25 +79,19 @@ def test_tutorial_01_overlay_falls_back_to_fixed_image_mask(
 class TestTutorial01HeartGatedCTToUSD:
     """End-to-end test for tutorial_01_heart_gated_ct_to_usd.py."""
 
-    _class_name = "tutorial_01_heart_gated_ct_to_usd"
+    _class_name = "tutorial_01"
 
     def test_run(self, test_directories: dict[str, Path]) -> None:
-        from tutorials.tutorial_01_heart_gated_ct_to_usd import run_tutorial
-
-        out_dir = test_directories["output"] / self._class_name
-        results: dict[str, Any] = run_tutorial(
-            data_dir=test_directories["data"],
-            output_dir=out_dir,
-            registration_method="ants",
-        )
+        out_dir = _REPO_ROOT / "tutorials" / "output" / "tutorial_01"
+        results = _run_tutorial_script("tutorial_01_heart_gated_ct_to_usd.py")
         assert results["usd_file"], "USD file path should not be empty"
         assert Path(results["usd_file"]).exists(), "USD file should exist"
+        assert results["screenshots"], "Tutorial 1 should produce screenshots"
 
         tt = TestTools(
             class_name=self._class_name,
-            results_dir=test_directories["output"],
-            baselines_dir=test_directories["baselines"],
-            results_output_dir=out_dir,
+            results_dir=out_dir,
+            baselines_dir=test_directories["baselines"] / self._class_name,
         )
         _compare_screenshots(results["screenshots"], tt)
 
@@ -213,21 +110,15 @@ class TestTutorial02CTToVTK:
     _class_name = "tutorial_02_ct_to_vtk"
 
     def test_run(self, test_directories: dict[str, Path]) -> None:
-        from tutorials.tutorial_02_ct_to_vtk import run_tutorial
-
-        out_dir = test_directories["output"] / self._class_name
-        results: dict[str, Any] = run_tutorial(
-            data_dir=test_directories["data"],
-            output_dir=out_dir,
-        )
+        out_dir = _REPO_ROOT / "tutorials" / "output" / "tutorial_02"
+        results = _run_tutorial_script("tutorial_02_ct_to_vtk.py")
         assert results["surface_file"].exists(), "Combined VTP surface should exist"
         assert results["mesh_file"].exists(), "Combined VTU mesh should exist"
 
         tt = TestTools(
             class_name=self._class_name,
-            results_dir=test_directories["output"],
-            baselines_dir=test_directories["baselines"],
-            results_output_dir=out_dir,
+            results_dir=out_dir,
+            baselines_dir=test_directories["baselines"] / self._class_name,
         )
         _compare_screenshots(results["screenshots"], tt)
 
@@ -252,44 +143,19 @@ def test_run(self, test_directories: dict[str, Path]) -> None:
                 "KCL-Heart-Model not downloaded. See data/README.md for instructions."
             )
 
-        from tutorials.tutorial_03_create_statistical_model import run_tutorial
-
-        out_dir = test_directories["output"] / self._class_name
-        results: dict[str, Any] = run_tutorial(
-            data_dir=test_directories["data"],
-            output_dir=out_dir,
-            pca_components=5,
-            max_samples=10,
-        )
+        out_dir = _REPO_ROOT / "tutorials" / "output" / "tutorial_03"
+        results = _run_tutorial_script("tutorial_03_create_statistical_model.py")
         assert results["model_file"].exists(), "pca_model.json should exist"
         assert results["mean_surface_file"].exists(), "Mean surface VTP should exist"
 
         tt = TestTools(
             class_name=self._class_name,
-            results_dir=test_directories["output"],
-            baselines_dir=test_directories["baselines"],
-            results_output_dir=out_dir,
+            results_dir=out_dir,
+            baselines_dir=test_directories["baselines"] / self._class_name,
         )
         _compare_screenshots(results["screenshots"], tt)
 
 
-# -----------------------------------------------------------------------------
-# Tutorial 4 - Fit Statistical Model to Patient
-# -----------------------------------------------------------------------------
-
-
-def test_tutorial_04_extract_surface_uses_dataset_surface() -> None:
-    """Use the robust dataset_surface algorithm for VTK surface extraction."""
-    from tutorials.tutorial_04_fit_statistical_model_to_patient import (
-        _extract_surface,
-    )
-
-    mesh: Any = _FakeSurfaceExtractable()
-
-    assert _extract_surface(mesh) is mesh
-    assert mesh.algorithm == "dataset_surface"
-
-
 @pytest.mark.tutorial
 @pytest.mark.requires_data
 @pytest.mark.slow
@@ -306,41 +172,25 @@ def test_run(self, test_directories: dict[str, Path]) -> None:
             )
 
         pca_json = (
-            test_directories["output"]
-            / "tutorial_03_create_statistical_model"
-            / "pca_model.json"
+            _REPO_ROOT / "tutorials" / "output" / "tutorial_03" / "pca_model.json"
         )
         if not pca_json.exists():
-            from tutorials.tutorial_03_create_statistical_model import (
-                run_tutorial as run_tutorial_03,
-            )
-
-            run_tutorial_03(
-                data_dir=test_directories["data"],
-                output_dir=pca_json.parent,
-                pca_components=5,
-                max_samples=10,
-            )
+            _run_tutorial_script("tutorial_03_create_statistical_model.py")
             assert pca_json.exists(), (
                 "Tutorial 3 bootstrap did not create the expected PCA model file: "
                 f"{pca_json}"
             )
 
-        from tutorials.tutorial_04_fit_statistical_model_to_patient import run_tutorial
-
-        out_dir = test_directories["output"] / self._class_name
-        results: dict[str, Any] = run_tutorial(
-            data_dir=test_directories["data"],
-            output_dir=out_dir,
-            pca_json=pca_json,
+        out_dir = _REPO_ROOT / "tutorials" / "output" / "tutorial_04"
+        results = _run_tutorial_script(
+            "tutorial_04_fit_statistical_model_to_patient.py"
         )
         assert results["registered_file"].exists(), "Registered VTP should exist"
 
         tt = TestTools(
             class_name=self._class_name,
-            results_dir=test_directories["output"],
-            baselines_dir=test_directories["baselines"],
-            results_output_dir=out_dir,
+            results_dir=out_dir,
+            baselines_dir=test_directories["baselines"] / self._class_name,
         )
         _compare_screenshots(results["screenshots"], tt)
 
@@ -361,9 +211,7 @@ class TestTutorial05VTKToUSD:
     def test_run(self, test_directories: dict[str, Path]) -> None:
         # Prefer Tutorial 2 output; fall back to any .vtp in data
         tutorial2_vtp = (
-            test_directories["output"]
-            / "tutorial_02_ct_to_vtk"
-            / "patient_surfaces.vtp"
+            _REPO_ROOT / "tutorials" / "output" / "tutorial_02" / "patient_surfaces.vtp"
         )
         vtk_file = tutorial2_vtp if tutorial2_vtp.exists() else None
         if vtk_file is None:
@@ -375,22 +223,15 @@ def test_run(self, test_directories: dict[str, Path]) -> None:
                 )
             vtk_file = found[0]
 
-        from tutorials.tutorial_05_vtk_to_usd import run_tutorial
-
-        out_dir = test_directories["output"] / self._class_name
-        results: dict[str, Any] = run_tutorial(
-            data_dir=test_directories["data"],
-            output_dir=out_dir,
-            vtk_file=vtk_file,
-        )
+        out_dir = _REPO_ROOT / "tutorials" / "output" / "tutorial_05"
+        results = _run_tutorial_script("tutorial_05_vtk_to_usd.py")
         assert results["usd_file"], "USD file path should not be empty"
         assert Path(results["usd_file"]).exists(), "USD file should exist"
 
         tt = TestTools(
             class_name=self._class_name,
-            results_dir=test_directories["output"],
-            baselines_dir=test_directories["baselines"],
-            results_output_dir=out_dir,
+            results_dir=out_dir,
+            baselines_dir=test_directories["baselines"] / self._class_name,
         )
         _compare_screenshots(results["screenshots"], tt)
 
@@ -415,16 +256,8 @@ def test_run(self, test_directories: dict[str, Path]) -> None:
                 "DirLab-4DCT Case1 not downloaded. See data/README.md for instructions."
             )
 
-        from tutorials.tutorial_06_reconstruct_highres_4d_ct import run_tutorial
-
-        out_dir = test_directories["output"] / self._class_name
-        results: dict[str, Any] = run_tutorial(
-            data_dir=test_directories["data"],
-            output_dir=out_dir,
-            case=1,
-            max_frames=3,
-            registration_method="ants",
-        )
+        out_dir = _REPO_ROOT / "tutorials" / "output" / "tutorial_06"
+        results = _run_tutorial_script("tutorial_06_reconstruct_highres_4d_ct.py")
         assert results["reconstructed_files"], (
             "At least one reconstructed frame expected"
         )
@@ -433,8 +266,7 @@ def test_run(self, test_directories: dict[str, Path]) -> None:
 
         tt = TestTools(
             class_name=self._class_name,
-            results_dir=test_directories["output"],
-            baselines_dir=test_directories["baselines"],
-            results_output_dir=out_dir,
+            results_dir=out_dir,
+            baselines_dir=test_directories["baselines"] / self._class_name,
         )
         _compare_screenshots(results["screenshots"], tt)
diff --git a/tests/test_vtk_to_usd_library.py b/tests/test_vtk_to_usd_library.py
index 0c9268d..399bffc 100644
--- a/tests/test_vtk_to_usd_library.py
+++ b/tests/test_vtk_to_usd_library.py
@@ -11,14 +11,17 @@
 - CHOP-Valve4D: data/CHOP-Valve4D/
 """
 
+import sys
 from pathlib import Path
 
 import numpy as np
 import pytest
 import pyvista as pv
-from pxr import UsdGeom, UsdShade
+from pxr import Gf, Usd, UsdGeom, UsdShade
 
 from physiomotion4d import ConvertVTKToUSD
+from physiomotion4d.test_tools import TestTools
+from physiomotion4d.usd_tools import USDTools
 
 
 def get_data_dir() -> Path:
@@ -122,6 +125,46 @@ def test_from_files_single_file_writes_static_mesh(self, tmp_path: Path) -> None
         assert len(mesh.GetPointsAttr().Get()) == plane.n_points
         assert not stage.HasAuthoredTimeCodeRange()
 
+    def test_openusd_screenshot_uses_vtk_loader(self, tmp_path: Path) -> None:
+        """Render a tiny OpenUSD mesh through TestTools without USD imaging plugins."""
+        usd_path = tmp_path / "triangle.usd"
+        stage = Usd.Stage.CreateNew(str(usd_path))
+        world = stage.DefinePrim("/World", "Xform")
+        stage.SetDefaultPrim(world)
+        mesh = UsdGeom.Mesh.Define(stage, "/World/Triangle")
+        mesh.CreatePointsAttr(
+            [
+                Gf.Vec3f(0.0, 0.0, 0.0),
+                Gf.Vec3f(1.0, 0.0, 0.0),
+                Gf.Vec3f(0.0, 1.0, 0.0),
+            ]
+        )
+        mesh.CreateFaceVertexCountsAttr([3])
+        mesh.CreateFaceVertexIndicesAttr([0, 1, 2])
+        stage.Save()
+
+        loaded = USDTools().load_usd_as_vtk(usd_path)
+        assert loaded.n_points == 3
+        assert "openusd_rgb" in loaded.point_data
+        assert np.all(loaded.point_data["openusd_rgb"] == np.array([255, 0, 0]))
+
+        # GitHub's hosted windows-latest runners lack a real OpenGL stack,
+        # so VTK's render path access-violates inside Plotter.screenshot
+        # with no Mesa software-render fallback available. Linux runners
+        # render off-screen successfully via xvfb + libosmesa6. Skip only
+        # on Windows; the load_usd_as_vtk assertions above still run.
+        if sys.platform == "win32":
+            pytest.skip(
+                "Skipping VTK render on Windows hosted runners: no usable "
+                "OpenGL context"
+            )
+
+        tt = TestTools(class_name="openusd", results_dir=tmp_path)
+        screenshot = tt.save_screenshot_openusd(usd_path, "triangle.png")
+
+        assert screenshot.exists()
+        assert screenshot.stat().st_size > 0
+
     def test_from_files_static_merge_writes_separate_meshes(
         self, tmp_path: Path
     ) -> None:
diff --git a/tutorials/README.md b/tutorials/README.md
index 1e90d2e..56c950f 100644
--- a/tutorials/README.md
+++ b/tutorials/README.md
@@ -11,28 +11,42 @@ dataset licensing, and expected directory layout.
 
 ## Tutorial Index
 
-| # | Script | Workflow Class | CLI Command | Dataset |
-|---|--------|---------------|-------------|---------|
-| 1 | [tutorial_01_heart_gated_ct_to_usd.py](tutorial_01_heart_gated_ct_to_usd.py) | `WorkflowConvertHeartGatedCTToUSD` | `physiomotion4d-heart-gated-ct` | Slicer-Heart-CT (prepare first) |
-| 2 | [tutorial_02_ct_to_vtk.py](tutorial_02_ct_to_vtk.py) | `WorkflowConvertCTToVTK` | `physiomotion4d-convert-ct-to-vtk` | Slicer-Heart-CT (prepare first) |
-| 3 | [tutorial_03_create_statistical_model.py](tutorial_03_create_statistical_model.py) | `WorkflowCreateStatisticalModel` | `physiomotion4d-create-statistical-model` | KCL-Heart-Model (manual) |
-| 4 | [tutorial_04_fit_statistical_model_to_patient.py](tutorial_04_fit_statistical_model_to_patient.py) | `WorkflowFitStatisticalModelToPatient` | `physiomotion4d-fit-statistical-model-to-patient` | KCL-Heart-Model plus Tutorial 3 output |
-| 5 | [tutorial_05_vtk_to_usd.py](tutorial_05_vtk_to_usd.py) | `WorkflowConvertVTKToUSD` | `physiomotion4d-convert-vtk-to-usd` | Output of tutorial 2 |
-| 6 | [tutorial_06_reconstruct_highres_4d_ct.py](tutorial_06_reconstruct_highres_4d_ct.py) | `WorkflowReconstructHighres4DCT` | `physiomotion4d-reconstruct-highres-4d-ct` | DirLab-4DCT (manual) |
+| # | Script | Primary API | Dataset |
+|---|--------|-------------|---------|
+| 1 | [tutorial_01_heart_gated_ct_to_usd.py](tutorial_01_heart_gated_ct_to_usd.py) | `WorkflowConvertHeartGatedCTToUSD` | Slicer-Heart-CT (prepare first) |
+| 2 | [tutorial_02_ct_to_vtk.py](tutorial_02_ct_to_vtk.py) | `WorkflowConvertCTToVTK` | Slicer-Heart-CT (prepare first) |
+| 3 | [tutorial_03_create_statistical_model.py](tutorial_03_create_statistical_model.py) | `WorkflowCreateStatisticalModel` | KCL-Heart-Model (manual) |
+| 4 | [tutorial_04_fit_statistical_model_to_patient.py](tutorial_04_fit_statistical_model_to_patient.py) | `WorkflowFitStatisticalModelToPatient` | KCL-Heart-Model plus Tutorial 3 output |
+| 5 | [tutorial_05_vtk_to_usd.py](tutorial_05_vtk_to_usd.py) | `WorkflowConvertVTKToUSD` | Output of tutorial 2 |
+| 6 | [tutorial_06_reconstruct_highres_4d_ct.py](tutorial_06_reconstruct_highres_4d_ct.py) | `WorkflowReconstructHighres4DCT` | DirLab-4DCT (manual) |
+| 7 | [tutorial_07_dirlab_pca_model.py](tutorial_07_dirlab_pca_model.py) | `WorkflowCreateStatisticalModel`, `WorkflowFitStatisticalModelToPatient` | DirLab-4DCT (manual) |
+| 8 | [tutorial_08_dirlab_pca_time_series.py](tutorial_08_dirlab_pca_time_series.py) | `RegisterTimeSeriesImages` | DirLab-4DCT plus Tutorial 7 output |
+| 9 | [tutorial_09_physicsnemo_mesh_stage_model.py](tutorial_09_physicsnemo_mesh_stage_model.py) | `physicsnemo.models.mlp.FullyConnected` | Tutorial 8 output |
 
 ## Running a Tutorial
 
-Each tutorial is a standalone Python script with a `run_tutorial(data_dir, output_dir)`
-function. Run from the repository root:
+Each tutorial is a standalone percent-cell Python script (`# %%`) that can be
+run cell-by-cell in VS Code or Cursor, or executed end-to-end as a regular
+Python script. Paths are defined near the top of each script. By default, data
+is read from the repository `data/` directory and outputs are written under
+`tutorials/output/<tutorial_name>/`.
 
 ```bash
-python tutorials/tutorial_01_heart_gated_ct_to_usd.py \
-    --data-dir ./data --output-dir ./output
-
-python tutorials/tutorial_02_ct_to_vtk.py \
-    --data-dir ./data --output-dir ./output
+# Run the whole tutorial from the command line
+python tutorials/tutorial_01_heart_gated_ct_to_usd.py
 ```
 
+In VS Code or Cursor, open the tutorial and use **Run Python File** (or run
+the cells in order with **Run Cell**). The script's `if __name__ ==
+"__main__":` block executes the workflow and assigns the resulting
+`tutorial_results` dict in the script's namespace; the same variable is what
+`tests/test_tutorials.py` consumes via `runpy.run_path(..., run_name=
+"__main__")`.
+
+To use different paths, edit the constants near the top of the tutorial
+script. For repeatable command-line execution with path arguments, use the
+installed `physiomotion4d-*` CLI commands instead.
+
 ## Running as Pytest Tutorial Tests
 
 All tutorials are wired into the test suite under the `tutorial` marker.
@@ -56,6 +70,9 @@ pytest tests/test_tutorials.py::TestTutorial01HeartGatedCTToUSD --run-tutorials
 3. **Tutorial 3** creates the PCA statistical model from KCL-Heart-Model.
 4. **Tutorial 4** applies the statistical model, consuming Tutorial 3 output.
 5. **Tutorial 6** requires DirLab-4DCT - download it per `data/README.md`.
+6. **Tutorial 7** creates a surface PCA model of the five lung lobes from DirLab-4DCT, then fits it to every available case.
+7. **Tutorial 8** registers DirLab respiratory phases with ANTs+ICON and propagates the Tutorial 7 fitted meshes through each time series.
+8. **Tutorial 9** trains a PhysicsNeMo model to predict a PCA-fitted mesh at a user-specified respiratory stage.
 
 ## For Contributors
 
diff --git a/tutorials/tutorial_01_heart_gated_ct_to_usd.py b/tutorials/tutorial_01_heart_gated_ct_to_usd.py
index f42700d..ebabe03 100644
--- a/tutorials/tutorial_01_heart_gated_ct_to_usd.py
+++ b/tutorials/tutorial_01_heart_gated_ct_to_usd.py
@@ -56,21 +56,6 @@
 - USDAnatomyTools (usd_anatomy_tools.py):
     Applies clinical material colours to USD prims (used internally).
 
-CLI Equivalent
---------------
-The same main outputs (without screenshots) can be produced via the CLI::
-
-    physiomotion4d-heart-gated-ct \\
-        data/Slicer-Heart-CT/TruncalValve_4DCT.seq.nrrd \\
-        --contrast \\
-        --project-name cardiac_model \\
-        --registration-method ants \\
-        --registration-iterations 1 \\
-        --output-dir ./output/tutorial_01
-
-See ``src/physiomotion4d/cli/convert_heart_gated_ct_to_usd.py`` for full CLI
-documentation.
-
 Data Required
 -------------
 See data/README.md for download instructions and dataset licensing.
@@ -80,190 +65,104 @@
 download notebook or download the file manually before running this tutorial.
 """
 
+# %%
+# Imports
 from __future__ import annotations
 
-import argparse
 import logging
 from pathlib import Path
-from typing import Any, Optional
 
 import itk
 
-from physiomotion4d.segment_chest_total_segmentator import SegmentChestTotalSegmentator
 from physiomotion4d.test_tools import TestTools
 from physiomotion4d.workflow_convert_heart_gated_ct_to_usd import (
     WorkflowConvertHeartGatedCTToUSD,
 )
 
+# nnUNetv2 (used by TotalSegmentator inside WorkflowConvertHeartGatedCTToUSD)
+# spawns a multiprocessing.Pool. On Windows the spawn start method re-imports
+# this script in each child; without the __name__ == "__main__" guard around
+# the top-level work, that re-import fires workflow.process() again and
+# Python's spawn-cascade detector raises RuntimeError.
+if __name__ == "__main__":
+    # %%
+    # Data directory specification
+    REPO_ROOT = Path(__file__).resolve().parent.parent
+    TUTORIALS_DIR = Path(__file__).resolve().parent
+    DATA_DIR = REPO_ROOT / "data"
+    FULL_DATA_DIR = DATA_DIR / "Slicer-Heart-CT"
+    TEST_DATA_DIR = DATA_DIR / "test" / "slicer_heart_small"
+    OUTPUT_DIR = TUTORIALS_DIR / "output" / "tutorial_01"
+    REGISTRATION_METHOD = "ants"
+    LOG_LEVEL = logging.INFO
+
+    # %%
+    # Data reading
+    test_mode = TestTools.running_as_test()
+
+    data_dir = TEST_DATA_DIR if test_mode else FULL_DATA_DIR
+    output_dir = OUTPUT_DIR
+    registration_method = REGISTRATION_METHOD
+    log_level = LOG_LEVEL
 
-class _CachedLabelmapSegmenter(SegmentChestTotalSegmentator):
-    """Segmenter that returns a cached labelmap for tutorial test data."""
-
-    def __init__(self, labelmap: Any, log_level: int | str = logging.INFO):
-        super().__init__(log_level=log_level)
-        self._labelmap = labelmap
-
-    def segmentation_method(self, preprocessed_image: Any) -> Any:
-        return self._labelmap
-
-
-def _first_current_contour_mesh(
-    workflow: WorkflowConvertHeartGatedCTToUSD,
-) -> Optional[Any]:
-    """Return a contour mesh produced by the current workflow run.
-
-    Prefer transformed all-anatomy contours, because those are the meshes passed
-    into USD conversion for the current run. Fall back to reference contours only
-    if transformed contours are unavailable.
-    """
-    transformed_contours = getattr(workflow, "_transformed_contours", {})
-    if isinstance(transformed_contours, dict):
-        for mesh in transformed_contours.get("all", []):
-            if getattr(mesh, "n_points", 0) > 0:
-                return mesh
-
-    reference_contours = getattr(workflow, "_reference_contours", {})
-    if isinstance(reference_contours, dict):
-        mesh = reference_contours.get("all")
-        if mesh is not None and getattr(mesh, "n_points", 0) > 0:
-            return mesh
-
-    return None
-
-
-def _current_reference_image(
-    workflow: WorkflowConvertHeartGatedCTToUSD,
-    output_dir: Path,
-) -> Optional[Any]:
-    """Return the reference image used by the current workflow run."""
-    fixed_image = getattr(workflow, "_fixed_image", None)
-    if fixed_image is not None:
-        return fixed_image
-
-    fixed_image_file = output_dir / "fixed_image.mha"
-    if fixed_image_file.exists():
-        return itk.imread(str(fixed_image_file))
-
-    ref_frames = sorted(output_dir.glob("slice_???.mha"))
-    if ref_frames:
-        return itk.imread(str(ref_frames[0]))
-
-    return None
-
-
-def _current_reference_segmentation(
-    workflow: WorkflowConvertHeartGatedCTToUSD,
-    output_dir: Path,
-) -> Optional[Any]:
-    """Return the labelmap for the current workflow reference image."""
-    fixed_segmentation = getattr(workflow, "_fixed_segmentation", None)
-    if isinstance(fixed_segmentation, dict):
-        labelmap = fixed_segmentation.get("labelmap")
-        if labelmap is not None:
-            return labelmap
-
-    fixed_mask_file = output_dir / "fixed_image_mask.mha"
-    if fixed_mask_file.exists():
-        return itk.imread(str(fixed_mask_file))
-
-    label_files = sorted(output_dir.glob("slice_???_labelmap*.mha"))
-    if label_files:
-        return itk.imread(str(label_files[0]))
-
-    return None
-
-
-def run_tutorial(
-    data_dir: Path,
-    output_dir: Path,
-    *,
-    registration_method: str = "ants",
-    log_level: int = logging.INFO,
-) -> dict[str, Any]:
-    """Run Tutorial 1: Heart-Gated CT to Animated USD.
-
-    Args:
-        data_dir: Root of the ``data/`` directory (see data/README.md).
-        output_dir: Directory to write outputs and screenshots.
-        registration_method: ``'ants'`` (CPU-capable, default) or ``'icon'`` (GPU).
-        log_level: Python logging level.
-
-    Returns:
-        dict with keys:
-
-        - ``'usd_file'`` (str): path to the final painted USD.
-        - ``'screenshots'`` (list[Path]): paths to saved PNG screenshots.
-          PNGs are rendered from data produced by this invocation, not from
-          previously saved VTK/VTP files in ``output_dir``. Reference-frame
-          screenshots use the workflow's selected fixed image.
-    """
     output_dir.mkdir(parents=True, exist_ok=True)
 
-    test_frame_files = sorted(data_dir.glob("slice_???_sml.mha"))[:2]
-    cached_labelmap: Path | None = None
-    if test_frame_files:
-        input_filenames = [str(path) for path in test_frame_files]
-        reference_frame = int(len(test_frame_files) * 0.7)
-        fixed_frame = test_frame_files[reference_frame]
-        labelmap_path = fixed_frame.with_name(f"{fixed_frame.stem}_labelmap.mha")
-        if labelmap_path.exists():
-            cached_labelmap = labelmap_path
+    if test_mode:
+        number_of_registration_iterations = 1
     else:
-        input_filenames = []
+        number_of_registration_iterations = 10
 
-    nrrd_candidates = [
-        data_dir / "Slicer-Heart-CT" / "TruncalValve_4DCT.seq.nrrd",
-        data_dir / "TruncalValve_4DCT.seq.nrrd",
-    ]
-    if not input_filenames:
-        nrrd_file = next((path for path in nrrd_candidates if path.exists()), None)
-        if nrrd_file is not None:
-            input_filenames = [str(nrrd_file)]
-            cached_labelmap = data_dir / "slice_014_sml_labelmap.mha"
+    # %%
+    frame_files = sorted(data_dir.glob("slice_???.mha"))
+    if test_mode:
+        frame_files = frame_files[:2]
 
+    input_filenames = [str(path) for path in frame_files]
     if not input_filenames:
         raise FileNotFoundError(
             "Slicer-Heart-CT data not found. Checked:\n"
-            + "\n".join(f"  - {path}" for path in nrrd_candidates)
+            + f"  - {data_dir}"
             + "\n"
-            "See data/README.md for download instructions."
+            + "See data/README.md for download instructions."
         )
 
+    # %%
+    # Workflow initialization
     workflow = WorkflowConvertHeartGatedCTToUSD(
         input_filenames=input_filenames,
         contrast_enhanced=True,
         output_directory=str(output_dir),
         project_name="cardiac_model",
         registration_method=registration_method,
-        number_of_registration_iterations=1,
+        number_of_registration_iterations=number_of_registration_iterations,
         log_level=log_level,
+        save_registered_images=True,
+        save_registration_transforms=True,
+        save_labelmaps=True,
     )
-    if cached_labelmap is not None and cached_labelmap.exists():
-        workflow.segmenter = _CachedLabelmapSegmenter(
-            itk.imread(str(cached_labelmap)),
-            log_level=log_level,
-        )
 
+    # %%
+    # Workflow execution
     usd_file = output_dir / workflow.process()
 
-    # Screenshots
+    # %%
+    # Result saving
     tt = TestTools(
+        class_name="tutorial_01_heart_gated_ct_to_usd",
         results_dir=output_dir,
-        baselines_dir=output_dir / "baselines",
-        class_name="tutorial_01",
-        results_output_dir=output_dir,
         log_level=log_level,
     )
 
     screenshots: list[Path] = []
 
-    # Reference frame: use the workflow's selected fixed image for this run.
-    ref_image = _current_reference_image(workflow, output_dir)
-    if ref_image is not None:
+    test_image_num = int(0.7 * len(input_filenames))
+    test_image_path = output_dir / f"slice_{test_image_num:03d}_registered.mha"
+    if test_image_path.exists():
+        test_image = itk.imread(str(test_image_path))
         screenshots.append(
             tt.save_screenshot_image_slice(
-                ref_image,
-                "reference_frame_axial.png",
+                test_image,
+                f"slice_{test_image_num:03d}_registered_test.png",
                 axis=0,
                 slice_fraction=0.5,
                 colormap="gray",
@@ -272,66 +171,28 @@ def run_tutorial(
             )
         )
 
-        # Segmentation overlay: align with the selected fixed image.
-        overlay = _current_reference_segmentation(workflow, output_dir)
-        screenshots.append(
-            tt.save_screenshot_image_slice(
-                ref_image,
-                "segmentation_overlay.png",
-                axis=0,
-                slice_fraction=0.5,
-                colormap="gray",
-                vmin=-200,
-                vmax=600,
-                overlay_mask=overlay,
+        test_labelmap_path = output_dir / f"slice_{test_image_num:03d}_labelmap.mha"
+        if test_labelmap_path.exists():
+            test_labelmap = itk.imread(str(test_labelmap_path))
+            screenshots.append(
+                tt.save_screenshot_image_slice(
+                    test_image,
+                    f"slice_{test_image_num:03d}_labelmap_test.png",
+                    axis=0,
+                    slice_fraction=0.5,
+                    colormap="gray",
+                    vmin=-200,
+                    vmax=600,
+                    overlay_mask=test_labelmap,
+                )
             )
-        )
 
-    # 3-D contour view: render the current run's in-memory contours.
-    contour_mesh = _first_current_contour_mesh(workflow)
-    if contour_mesh is not None:
+    if usd_file.exists():
         screenshots.append(
-            tt.save_screenshot_mesh(
-                contour_mesh,
-                "contours_3d.png",
-                camera_position="iso",
-                color="tomato",
-                opacity=0.85,
+            tt.save_screenshot_openusd(
+                usd_file,
+                "cardiac_model_test.png",
             )
         )
 
-    return {"usd_file": str(usd_file), "screenshots": screenshots}
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser(
-        description=__doc__,
-        formatter_class=argparse.RawDescriptionHelpFormatter,
-    )
-    parser.add_argument(
-        "--data-dir",
-        type=Path,
-        default=Path("data"),
-        help="Root data directory (default: ./data)",
-    )
-    parser.add_argument(
-        "--output-dir",
-        type=Path,
-        default=Path("output") / "tutorial_01",
-        help="Output directory (default: ./output/tutorial_01)",
-    )
-    parser.add_argument(
-        "--registration-method",
-        default="ants",
-        choices=["ants", "icon"],
-        help="Registration method: ants (CPU) or icon (GPU). Default: ants",
-    )
-    args = parser.parse_args()
-
-    results = run_tutorial(
-        args.data_dir,
-        args.output_dir,
-        registration_method=args.registration_method,
-    )
-    print(f"USD file: {results['usd_file']}")
-    print(f"Screenshots: {[str(p) for p in results['screenshots']]}")
+    tutorial_results = {"usd_file": str(usd_file), "screenshots": screenshots}
diff --git a/tutorials/tutorial_02_ct_to_vtk.py b/tutorials/tutorial_02_ct_to_vtk.py
index 6e001fa..02c406a 100644
--- a/tutorials/tutorial_02_ct_to_vtk.py
+++ b/tutorials/tutorial_02_ct_to_vtk.py
@@ -3,202 +3,109 @@
 
 Purpose
 -------
-Segment a 3D CT image into anatomical groups (heart, lungs, vessels, bone,
-soft tissue) and export per-group VTK surface and mesh files. Each output mesh
-is annotated with anatomy metadata and colour so it can be used directly in
-PyVista, ParaView, or the downstream USD pipeline (Tutorial 5).
-
-Inputs
-------
-- A single 3D CT image in any ITK-readable format (NIfTI, MHA, NRRD, etc.).
-  This tutorial uses one time frame from the Slicer-Heart-CT dataset.
-  Expected location: ``data/Slicer-Heart-CT/`` (any ``slice_???.mha`` frame).
-
-Outputs
--------
-- ``output_dir/patient_surfaces.vtp`` - all anatomy surfaces in one file
-- ``output_dir/patient_meshes.vtu`` - all voxel meshes in one file
-- Screenshots (PNG):
-  - ``segmentation_overlay.png`` - segmentation mask overlaid on axial CT slice
-  - ``vtk_surfaces.png`` - 3-D isometric view of the combined surface
-
-Strengths
----------
-- One call to ``WorkflowConvertCTToVTK.run_workflow()`` handles segmentation and
-  mesh extraction for all anatomy groups in a single pass.
-- Each output mesh carries field data (group name, label IDs, colour) for
-  downstream tools.
-- Combined-file output (default) produces a single VTP/VTU rather than one file
-  per group, simplifying downstream handling.
-
-Weaknesses / Limitations
-------------------------
-- TotalSegmentator requires ~8 GB GPU VRAM for full segmentation; CPU fallback
-  is available but much slower (~30 min per volume).
-- Small or unusual anatomical structures (e.g., pediatric heart) may be partially
-  missed by the default TotalSegmentator model.
-- Output mesh resolution is governed by the input CT voxel size; coarse scans
-  yield coarser meshes.
-
-Classes Used
-------------
-- WorkflowConvertCTToVTK (workflow_convert_ct_to_vtk.py):
-    Segments a CT image and extracts per-anatomy-group VTK surfaces and meshes.
-- SegmentChestTotalSegmentator (segment_chest_total_segmentator.py):
-    Deep-learning segmentation backend (used internally).
-- ContourTools (contour_tools.py):
-    Mesh extraction via marching cubes (used internally).
-- USDAnatomyTools (usd_anatomy_tools.py):
-    Provides anatomy group colours for mesh annotation (used internally).
-
-CLI Equivalent
---------------
-The same main outputs (without screenshots) can be produced via the CLI::
-
-    physiomotion4d-convert-ct-to-vtk \\
-        --input-image data/Slicer-Heart-CT/slice_000.mha \\
-        --output-dir ./output/tutorial_02 \\
-        --output-prefix patient \\
-        --contrast
-
-See ``src/physiomotion4d/cli/convert_ct_to_vtk.py`` for full CLI documentation.
+Segment one 3D CT frame into anatomical groups and save combined VTK surface
+and voxel mesh files. The output can be inspected directly in PyVista or used
+as input for Tutorial 5.
 
 Data Required
 -------------
-See data/README.md for download instructions and dataset licensing.
-Dataset: Slicer-Heart-CT - https://github.com/Slicer-Heart-CT/Slicer-Heart-CT
-Auto-download: the conftest fixture downloads
-``data/test/TruncalValve_4DCT.seq.nrrd`` and extracts frames as
-``data/test/slice_???.mha``. For this tutorial the full dataset is at
-``data/Slicer-Heart-CT/``.
+Full data: ``data/Slicer-Heart-CT/slice_???.mha``
+Test data: ``data/test/slicer_heart_small/slice_???.mha``
 """
 
+# %%
+# Imports
 from __future__ import annotations
 
-import argparse
 import logging
 from pathlib import Path
-from typing import Any
 
 import itk
 import pyvista as pv
 
-from physiomotion4d.segment_chest_total_segmentator import SegmentChestTotalSegmentator
 from physiomotion4d.test_tools import TestTools
 from physiomotion4d.workflow_convert_ct_to_vtk import WorkflowConvertCTToVTK
 
+# nnUNetv2 (used by TotalSegmentator inside several workflows) spawns a
+# multiprocessing.Pool. On Windows the spawn start method re-imports this
+# script in each child; without the __name__ == "__main__" guard around
+# top-level work, that re-import fires the segmenter again and Python's
+# spawn-cascade detector raises RuntimeError. Wrapping consistently across
+# tutorials also matches the style of tutorial_01.
+if __name__ == "__main__":
+    # %%
+    # Data directory specification
+    REPO_ROOT = Path(__file__).resolve().parent.parent
+    TUTORIALS_DIR = Path(__file__).resolve().parent
+    DATA_DIR = REPO_ROOT / "data"
+    FULL_DATA_DIR = DATA_DIR / "Slicer-Heart-CT"
+    TEST_DATA_DIR = DATA_DIR / "test" / "slicer_heart_small"
+    OUTPUT_DIR = TUTORIALS_DIR / "output" / "tutorial_02"
+    BASELINES_DIR = REPO_ROOT / "tests" / "baselines"
+    LOG_LEVEL = logging.INFO
+
+    # %%
+    # Data reading
+    test_mode = TestTools.running_as_test()
+
+    data_dir = TEST_DATA_DIR if test_mode else FULL_DATA_DIR
+    output_dir = OUTPUT_DIR
+    log_level = LOG_LEVEL
 
-class _CachedLabelmapSegmenter(SegmentChestTotalSegmentator):
-    """Segmenter that returns a cached labelmap for tutorial test data."""
-
-    def __init__(self, labelmap: Any, log_level: int | str = logging.INFO):
-        super().__init__(log_level=log_level)
-        self._labelmap = labelmap
-
-    def segmentation_method(self, preprocessed_image: Any) -> Any:
-        return self._labelmap
-
-
-class _CachedLabelmapWorkflow(WorkflowConvertCTToVTK):
-    """Workflow variant that uses a cached labelmap instead of TotalSegmentator."""
-
-    def __init__(
-        self,
-        labelmap: Any,
-        *,
-        segmentation_method: str = "total_segmentator",
-        log_level: int | str = logging.INFO,
-    ) -> None:
-        super().__init__(
-            segmentation_method=segmentation_method,
-            log_level=log_level,
-        )
-        self._labelmap = labelmap
-
-    def _create_segmenter(self) -> SegmentChestTotalSegmentator:
-        return _CachedLabelmapSegmenter(self._labelmap, log_level=self.log_level)
-
-
-def run_tutorial(
-    data_dir: Path,
-    output_dir: Path,
-    *,
-    log_level: int = logging.INFO,
-) -> dict[str, Any]:
-    """Run Tutorial 2: CT Segmentation to VTK Surfaces.
-
-    Args:
-        data_dir: Root of the ``data/`` directory (see data/README.md).
-        output_dir: Directory to write outputs and screenshots.
-        log_level: Python logging level.
-
-    Returns:
-        dict with keys:
-
-        - ``'result'`` (dict): workflow result dict with ``'surfaces'`` and
-          ``'meshes'`` entries (pv.PolyData / pv.UnstructuredGrid per group).
-        - ``'surface_file'`` (Path): path to the combined ``.vtp`` file.
-        - ``'mesh_file'`` (Path): path to the combined ``.vtu`` file.
-        - ``'screenshots'`` (list[Path]): paths to saved PNG screenshots.
-    """
     output_dir.mkdir(parents=True, exist_ok=True)
 
-    # Prefer full dataset; fall back to direct test-cache layouts.
-    candidates = list((data_dir / "Slicer-Heart-CT").glob("slice_???.mha"))
-    if not candidates:
-        candidates = list(data_dir.glob("slice_???_sml.mha"))
-    if not candidates:
-        candidates = list((data_dir / "test").glob("slice_???_sml.mha"))
-    if not candidates:
+    frame_files = sorted(data_dir.glob("slice_???.mha"))
+    if not frame_files:
         raise FileNotFoundError(
-            "No CT frame found under data/Slicer-Heart-CT/, data/test/, "
-            "or the provided data directory.\n"
-            "See data/README.md for download instructions."
+            "Slicer-Heart-CT frame data not found. Checked:\n"
+            + f"  - {data_dir}\n"
+            + "See data/README.md for download instructions."
         )
-    candidates.sort()
-    ct_file = candidates[0]
 
+    ct_file = frame_files[0]
     ct_image = itk.imread(str(ct_file))
 
-    labelmap_file = ct_file.with_name(f"{ct_file.stem}_labelmap.mha")
-    if labelmap_file.exists():
-        workflow = _CachedLabelmapWorkflow(
-            itk.imread(str(labelmap_file)),
-            segmentation_method="total_segmentator",
-            log_level=log_level,
-        )
-    else:
-        workflow = WorkflowConvertCTToVTK(
-            segmentation_method="total_segmentator",
-            log_level=log_level,
-        )
+    # %%
+    # Workflow initialization
+    workflow = WorkflowConvertCTToVTK(
+        segmentation_method="total_segmentator",
+        log_level=log_level,
+    )
+
+    # %%
+    # Workflow execution
     result = workflow.run_workflow(
         input_image=ct_image,
         contrast_enhanced_study=True,
     )
 
-    surface_file = output_dir / "patient_surfaces.vtp"
-    mesh_file = output_dir / "patient_meshes.vtu"
-    WorkflowConvertCTToVTK.save_combined_surface(
-        result["surfaces"], str(output_dir), prefix="patient"
+    # %%
+    # Result saving
+    surface_file = Path(
+        WorkflowConvertCTToVTK.save_combined_surface(
+            result["surfaces"],
+            str(output_dir),
+            prefix="patient",
+        )
     )
-    WorkflowConvertCTToVTK.save_combined_mesh(
-        result["meshes"], str(output_dir), prefix="patient"
+    mesh_file = Path(
+        WorkflowConvertCTToVTK.save_combined_mesh(
+            result["meshes"],
+            str(output_dir),
+            prefix="patient",
+        )
     )
+    labelmap_file = output_dir / "patient_labelmap.mha"
+    itk.imwrite(result["labelmap"], str(labelmap_file), compression=True)
 
-    # Screenshots
     tt = TestTools(
+        class_name="tutorial_02_ct_to_vtk",
         results_dir=output_dir,
-        baselines_dir=output_dir / "baselines",
-        class_name="tutorial_02",
-        results_output_dir=output_dir,
+        baselines_dir=BASELINES_DIR,
         log_level=log_level,
     )
 
     screenshots: list[Path] = []
-
-    # Segmentation overlay on axial CT slice
-    labelmap = result.get("labelmap")
     screenshots.append(
         tt.save_screenshot_image_slice(
             ct_image,
@@ -208,17 +115,18 @@ def run_tutorial(
             colormap="gray",
             vmin=-200,
             vmax=600,
-            overlay_mask=labelmap,
+            overlay_mask=result["labelmap"],
         )
     )
 
-    # 3-D view of combined surface
-    surfaces = [s for s in result["surfaces"].values() if s is not None]
+    surfaces = [
+        surface for surface in result["surfaces"].values() if surface is not None
+    ]
     if surfaces:
-        combined = pv.merge(surfaces) if len(surfaces) > 1 else surfaces[0]
+        combined_surface = pv.merge(surfaces) if len(surfaces) > 1 else surfaces[0]
         screenshots.append(
             tt.save_screenshot_mesh(
-                combined,
+                combined_surface,
                 "vtk_surfaces.png",
                 camera_position="iso",
                 color="lightblue",
@@ -226,34 +134,10 @@ def run_tutorial(
             )
         )
 
-    return {
+    tutorial_results = {
         "result": result,
         "surface_file": surface_file,
         "mesh_file": mesh_file,
+        "labelmap_file": labelmap_file,
         "screenshots": screenshots,
     }
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser(
-        description=__doc__,
-        formatter_class=argparse.RawDescriptionHelpFormatter,
-    )
-    parser.add_argument(
-        "--data-dir",
-        type=Path,
-        default=Path("data"),
-        help="Root data directory (default: ./data)",
-    )
-    parser.add_argument(
-        "--output-dir",
-        type=Path,
-        default=Path("output") / "tutorial_02",
-        help="Output directory (default: ./output/tutorial_02)",
-    )
-    args = parser.parse_args()
-
-    results = run_tutorial(args.data_dir, args.output_dir)
-    print(f"Surface file: {results['surface_file']}")
-    print(f"Mesh file:    {results['mesh_file']}")
-    print(f"Screenshots:  {[str(p) for p in results['screenshots']]}")
diff --git a/tutorials/tutorial_03_create_statistical_model.py b/tutorials/tutorial_03_create_statistical_model.py
index f0451c9..22eb6d1 100644
--- a/tutorials/tutorial_03_create_statistical_model.py
+++ b/tutorials/tutorial_03_create_statistical_model.py
@@ -3,79 +3,19 @@
 
 Purpose
 -------
-Build a PCA (Principal Component Analysis) statistical shape model from a
-population of anatomical meshes aligned to a reference. The model captures
-the mean shape and the principal modes of geometric variation across the
-population. The resulting model can be used in Tutorial 4 to constrain
-patient-specific fitting to anatomically plausible shapes.
-
-Inputs
-------
-- A reference mesh (``pv.DataSet`` / ``.vtu``): defines the template topology.
-  Expected location: ``data/KCL-Heart-Model/pca_mean.vtu``
-- A collection of sample meshes (list of ``pv.DataSet`` / ``.vtu``):
-  population shapes to learn from.
-  Expected location: ``data/KCL-Heart-Model/sample_meshes/*.vtu``
-
-Outputs
--------
-- ``output_dir/pca_model.json`` - PCA model (components and eigenvalues)
-- ``output_dir/pca_mean_surface.vtp`` - mean shape as a surface
-- Screenshots (PNG):
-  - ``pca_mean_model.png`` - 3-D view of the PCA mean surface
-  - ``pca_mode_01.png`` - mean +/- 2 sigma for the first PCA mode (side-by-side)
-  - ``pca_mode_02.png`` - mean +/- 2 sigma for the second PCA mode
-
-Strengths
----------
-- Single call to ``WorkflowCreateStatisticalModel.run_workflow()`` covers the
-  full pipeline: ICP alignment, deformable correspondence, and PCA.
-- Returns a pure-Python dict (PCA components and eigenvalues) compatible with
-  ``WorkflowFitStatisticalModelToPatient.set_use_pca_registration()``.
-- Surface-mode PCA (``solve_for_surface_pca=True``, default) is faster and
-  sufficient for most cardiac applications.
-
-Weaknesses / Limitations
-------------------------
-- Requires the KCL-Heart-Model dataset (manual download; see data/README.md).
-- Population size directly affects model quality; small populations (<20 meshes)
-  produce unreliable high-order modes.
-- ICP alignment (step 2) assumes all sample meshes share a common approximate
-  orientation; large pose variation may degrade correspondence.
-- Deformable correspondence (step 3) uses ANTs, which is slow on CPU.
-
-Classes Used
-------------
-- WorkflowCreateStatisticalModel (workflow_create_statistical_model.py):
-    Runs the full pipeline: ICP -> deformable correspondence -> PCA.
-- RegisterModelsICP (register_models_icp.py):
-    Aligns each sample to the reference (used internally).
-- RegisterModelsDistanceMaps (register_models_distance_maps.py):
-    Dense deformable correspondence via signed distance maps (used internally).
-
-CLI Equivalent
---------------
-The same main outputs (without screenshots) can be produced via the CLI::
-
-    physiomotion4d-create-statistical-model \\
-        --sample-meshes-dir data/KCL-Heart-Model/sample_meshes \\
-        --reference-mesh data/KCL-Heart-Model/pca_mean.vtu \\
-        --pca-components 10 \\
-        --output-dir ./output/tutorial_03
-
-See ``src/physiomotion4d/cli/create_statistical_model.py`` for full CLI
-documentation.
+Build a PCA statistical shape model from a reference mesh and a small population
+of sample meshes. Tutorial 4 can reuse the saved ``pca_model.json``.
 
 Data Required
 -------------
-See data/README.md for download instructions and dataset licensing.
-Dataset: KCL-Heart-Model - manual download required.
-Place files under ``data/KCL-Heart-Model/`` as described in data/README.md.
+Full data: ``data/KCL-Heart-Model``
+Test data: ``data/test/KCL-Heart-Model``
 """
 
+# %%
+# Imports
 from __future__ import annotations
 
-import argparse
 import json
 import logging
 from pathlib import Path
@@ -89,95 +29,98 @@
     WorkflowCreateStatisticalModel,
 )
 
+# nnUNetv2 (used by TotalSegmentator inside several workflows) spawns a
+# multiprocessing.Pool. On Windows the spawn start method re-imports this
+# script in each child; without the __name__ == "__main__" guard around
+# top-level work, that re-import fires the segmenter again and Python's
+# spawn-cascade detector raises RuntimeError. Wrapping consistently across
+# tutorials also matches the style of tutorial_01.
+if __name__ == "__main__":
+    # %%
+    # Data directory specification
+    REPO_ROOT = Path(__file__).resolve().parent.parent
+    TUTORIALS_DIR = Path(__file__).resolve().parent
+    DATA_DIR = REPO_ROOT / "data"
+    FULL_DATA_DIR = DATA_DIR / "KCL-Heart-Model"
+    TEST_DATA_DIR = DATA_DIR / "test" / "KCL-Heart-Model"
+    OUTPUT_DIR = TUTORIALS_DIR / "output" / "tutorial_03"
+    BASELINES_DIR = REPO_ROOT / "tests" / "baselines"
+    PCA_COMPONENTS = 10
+    MAX_SAMPLES = 20
+    LOG_LEVEL = logging.INFO
+
+    # %%
+    # Data reading
+    test_mode = TestTools.running_as_test()
+
+    data_dir = TEST_DATA_DIR if test_mode else FULL_DATA_DIR
+    output_dir = OUTPUT_DIR
+    log_level = LOG_LEVEL
+
+    if test_mode:
+        pca_components = min(PCA_COMPONENTS, 5)
+        max_samples = min(MAX_SAMPLES, 10)
+    else:
+        pca_components = PCA_COMPONENTS
+        max_samples = MAX_SAMPLES
 
-def run_tutorial(
-    data_dir: Path,
-    output_dir: Path,
-    *,
-    pca_components: int = 10,
-    max_samples: int = 20,
-    log_level: int = logging.INFO,
-) -> dict[str, Any]:
-    """Run Tutorial 3: Create a PCA Statistical Shape Model.
-
-    Args:
-        data_dir: Root of the ``data/`` directory (see data/README.md).
-        output_dir: Directory to write outputs and screenshots.
-        pca_components: Number of PCA modes to retain.
-        max_samples: Maximum number of sample meshes to use (cap for speed).
-        log_level: Python logging level.
-
-    Returns:
-        dict with keys:
-
-        - ``'pca_model'`` (dict): PCA model dict (components and eigenvalues).
-        - ``'mean_surface'`` (pv.PolyData): mean shape surface.
-        - ``'model_file'`` (Path): path to saved ``pca_model.json``.
-        - ``'mean_surface_file'`` (Path): path to saved ``pca_mean_surface.vtp``.
-        - ``'screenshots'`` (list[Path]): paths to saved PNG screenshots.
-    """
     output_dir.mkdir(parents=True, exist_ok=True)
 
-    kcl_dir = data_dir / "KCL-Heart-Model"
-    reference_file = kcl_dir / "pca_mean.vtu"
+    reference_file = data_dir / "pca_mean.vtu"
     if not reference_file.exists():
         raise FileNotFoundError(
             f"KCL-Heart-Model reference mesh not found: {reference_file}\n"
-            "See data/README.md for manual download instructions."
+            "See data/README.md for download instructions."
         )
 
-    sample_dir = kcl_dir / "sample_meshes"
+    sample_dir = data_dir / "sample_meshes"
     sample_files = sorted(sample_dir.glob("*.vtu"))
     if not sample_files:
-        sample_files = sorted(kcl_dir.glob("*.vtu"))
-    sample_files = [f for f in sample_files if f.name != "pca_mean.vtu"]
+        sample_files = sorted(data_dir.glob("*.vtu"))
+    sample_files = [path for path in sample_files if path.name != "pca_mean.vtu"]
     sample_files = sample_files[:max_samples]
     if len(sample_files) < 3:
         raise FileNotFoundError(
-            f"Need at least 3 non-reference sample meshes under {sample_dir} "
-            f"or {kcl_dir}.\n"
-            "See data/README.md for manual download instructions."
+            f"Need at least 3 sample meshes under {sample_dir} or {data_dir}.\n"
+            "See data/README.md for download instructions."
         )
 
     reference_mesh = cast(pv.DataSet, pv.read(str(reference_file)))
-    sample_meshes = [cast(pv.DataSet, pv.read(str(f))) for f in sample_files]
+    sample_meshes = [cast(pv.DataSet, pv.read(str(path))) for path in sample_files]
 
+    # %%
+    # Workflow initialization
     workflow = WorkflowCreateStatisticalModel(
         sample_meshes=sample_meshes,
         reference_mesh=reference_mesh,
         pca_number_of_components=pca_components,
         log_level=log_level,
     )
-    result = workflow.run_workflow()
 
-    mean_surface: pv.PolyData = result["mean_surface"]
-    mean_surface_file = output_dir / "pca_mean_surface.vtp"
-    mean_surface.save(str(mean_surface_file))
+    # %%
+    # Workflow execution
+    result = workflow.run_workflow()
 
-    # Serialise the JSON-safe parts of the PCA model
+    # %%
+    # Result saving
     pca_model: dict[str, Any] = result["pca_model"]
+    mean_surface: pv.PolyData = result["pca_mean_surface"]
+
     model_file = output_dir / "pca_model.json"
-    json_safe: dict[str, Any] = {}
-    for k, v in pca_model.items():
-        if isinstance(v, np.ndarray):
-            json_safe[k] = v.tolist()
-        elif isinstance(v, (int, float, str, bool, list)):
-            json_safe[k] = v
-    with open(model_file, "w") as fh:
-        json.dump(json_safe, fh, indent=2)
+    with model_file.open("w", encoding="utf-8") as f:
+        json.dump(pca_model, f, indent=2)
+
+    mean_surface_file = output_dir / "pca_mean_surface.vtp"
+    mean_surface.save(str(mean_surface_file))
 
-    # Screenshots
     tt = TestTools(
+        class_name="tutorial_03_create_statistical_model",
         results_dir=output_dir,
-        baselines_dir=output_dir / "baselines",
-        class_name="tutorial_03",
-        results_output_dir=output_dir,
+        baselines_dir=BASELINES_DIR,
         log_level=log_level,
     )
 
     screenshots: list[Path] = []
-
-    # Mean model
     screenshots.append(
         tt.save_screenshot_mesh(
             mean_surface,
@@ -188,90 +131,45 @@ def run_tutorial(
         )
     )
 
-    # First two PCA modes: show mean +/- 2 sigma side-by-side
-    components: Any = pca_model.get("components")
-    eigenvalues: Any = pca_model.get("eigenvalues")
+    components = pca_model.get("components", [])
+    eigenvalues = pca_model.get("eigenvalues", [])
     mean_points = np.asarray(mean_surface.points)
+    mode_count = min(2, pca_components, len(components), len(eigenvalues))
 
     try:
         pv.start_xvfb()
     except Exception:
         pass
 
-    if components is None or eigenvalues is None:
-        mode_count = 0
-    else:
-        mode_count = min(2, pca_components, len(components), len(eigenvalues))
-
     for mode_idx in range(mode_count):
         sigma = float(np.sqrt(eigenvalues[mode_idx]))
-        ev = np.asarray(components[mode_idx]).reshape(-1, 3)
+        mode_offsets = np.asarray(components[mode_idx]).reshape(-1, 3)
 
         minus_mesh = mean_surface.copy()
-        minus_mesh.points = mean_points - 2 * sigma * ev
+        minus_mesh.points = mean_points - 2.0 * sigma * mode_offsets
         plus_mesh = mean_surface.copy()
-        plus_mesh.points = mean_points + 2 * sigma * ev
+        plus_mesh.points = mean_points + 2.0 * sigma * mode_offsets
 
         plotter = pv.Plotter(off_screen=True, window_size=[1200, 500], shape=(1, 3))
         plotter.subplot(0, 0)
         plotter.add_mesh(minus_mesh, color="royalblue", opacity=0.9)
-        plotter.add_text("mean - 2 sigma", font_size=10)
         plotter.camera_position = "iso"
         plotter.subplot(0, 1)
         plotter.add_mesh(mean_surface, color="steelblue", opacity=0.9)
-        plotter.add_text("mean", font_size=10)
         plotter.camera_position = "iso"
         plotter.subplot(0, 2)
         plotter.add_mesh(plus_mesh, color="coral", opacity=0.9)
-        plotter.add_text("mean + 2 sigma", font_size=10)
         plotter.camera_position = "iso"
 
-        png_name = f"pca_mode_{mode_idx + 1:02d}.png"
-        png_path = output_dir / png_name
-        png_path.parent.mkdir(parents=True, exist_ok=True)
+        png_path = output_dir / f"pca_mode_{mode_idx + 1:02d}.png"
         plotter.screenshot(str(png_path))
         plotter.close()
         screenshots.append(png_path)
 
-    return {
+    tutorial_results = {
         "pca_model": pca_model,
         "mean_surface": mean_surface,
         "model_file": model_file,
         "mean_surface_file": mean_surface_file,
         "screenshots": screenshots,
     }
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser(
-        description=__doc__,
-        formatter_class=argparse.RawDescriptionHelpFormatter,
-    )
-    parser.add_argument(
-        "--data-dir",
-        type=Path,
-        default=Path("data"),
-        help="Root data directory (default: ./data)",
-    )
-    parser.add_argument(
-        "--output-dir",
-        type=Path,
-        default=Path("output") / "tutorial_03",
-        help="Output directory (default: ./output/tutorial_03)",
-    )
-    parser.add_argument(
-        "--pca-components",
-        type=int,
-        default=10,
-        help="Number of PCA modes to retain (default: 10)",
-    )
-    args = parser.parse_args()
-
-    results = run_tutorial(
-        args.data_dir,
-        args.output_dir,
-        pca_components=args.pca_components,
-    )
-    print(f"PCA model:    {results['model_file']}")
-    print(f"Mean surface: {results['mean_surface_file']}")
-    print(f"Screenshots:  {[str(p) for p in results['screenshots']]}")
diff --git a/tutorials/tutorial_04_fit_statistical_model_to_patient.py b/tutorials/tutorial_04_fit_statistical_model_to_patient.py
index 3f44820..d9d4c3a 100644
--- a/tutorials/tutorial_04_fit_statistical_model_to_patient.py
+++ b/tutorials/tutorial_04_fit_statistical_model_to_patient.py
@@ -3,245 +3,166 @@
 
 Purpose
 -------
-Register a generic anatomical template (e.g., a statistical heart model) to
-patient-specific surface meshes derived from medical imaging. The multi-stage
-pipeline performs ICP rough alignment, optional PCA-constrained shape fitting,
-mask-based deformable registration, and optional image-based refinement. The
-result is a patient-specific instance of the template mesh that matches the
-patient anatomy.
-
-Inputs
-------
-- Template model (``pv.UnstructuredGrid`` / ``.vtu``): generic anatomical mesh,
-  e.g., the KCL heart model.
-  Expected location: ``data/KCL-Heart-Model/pca_mean.vtu``
-- Patient surface models (list of ``pv.PolyData`` / ``.vtp``): anatomy surfaces
-  extracted from the patient CT (e.g., from Tutorial 2).
-  Expected location: ``data/KCL-Heart-Model/sample_meshes/*.vtu`` (used as
-  stand-in patient models for demonstration).
-- Optional patient CT image (``itk.Image``): used for image-based refinement.
-
-Outputs
--------
-- ``output_dir/registered_template.vtp`` - template mesh fitted to patient
-- Screenshots (PNG):
-  - ``model_before_registration.png`` - template and patient overlaid (pre-ICP)
-  - ``model_after_registration.png`` - registered template on patient
-
-Strengths
----------
-- Combines three complementary registration strategies in sequence, each
-  correcting different scales of misalignment.
-- PCA-constrained fitting (optional) prevents anatomically implausible
-  deformations by constraining shape variation to the training population.
-- Automatic mask generation means patient meshes are the only required input;
-  a CT image is optional.
-
-Weaknesses / Limitations
-------------------------
-- Requires the KCL-Heart-Model dataset (manual download; see data/README.md).
-- Deformable registration (ANTs) is the slowest stage (~5-15 min on CPU).
-- PCA mode is only beneficial when the template was trained on a population
-  that includes the patient's anatomical variant.
-- ICON-based image refinement requires a GPU.
-
-Classes Used
-------------
-- WorkflowFitStatisticalModelToPatient (workflow_fit_statistical_model_to_patient.py):
-    Orchestrates ICP -> (optional PCA) -> mask-to-mask -> (optional image) pipeline.
-- RegisterModelsICP (register_models_icp.py):
-    Centroid alignment followed by ICP affine registration (used internally).
-- RegisterModelsDistanceMaps (register_models_distance_maps.py):
-    ANTs deformable registration via signed distance maps (used internally).
-- ContourTools (contour_tools.py):
-    Creates reference images and masks from meshes (used internally).
-
-CLI Equivalent
---------------
-The same main outputs (without screenshots) can be produced via the CLI::
-
-    physiomotion4d-fit-statistical-model-to-patient \\
-        --template-model data/KCL-Heart-Model/pca_mean.vtu \\
-        --patient-models data/KCL-Heart-Model/sample_meshes/sample_000.vtu \\
-        --pca-json ./output/tutorial_03/pca_model.json \\
-        --output-dir ./output/tutorial_04
-
-See ``src/physiomotion4d/cli/fit_statistical_model_to_patient.py`` for full
-CLI documentation.
+Fit a generic anatomical template mesh to one or more patient-like surface
+meshes. If Tutorial 3 has already written ``pca_model.json``, the workflow uses
+that model to constrain the fitted shape.
 
 Data Required
 -------------
-See data/README.md for download instructions and dataset licensing.
-Dataset: KCL-Heart-Model - manual download required.
-Place files under ``data/KCL-Heart-Model/`` as described in data/README.md.
+Full data: ``data/KCL-Heart-Model``
+Test data: ``data/test/KCL-Heart-Model``
 """
 
+# %%
+# Imports
 from __future__ import annotations
 
-import argparse
 import json
 import logging
 from pathlib import Path
-from typing import Any, Optional, cast
+from typing import Any, cast
 
 import pyvista as pv
 
+from physiomotion4d.test_tools import TestTools
 from physiomotion4d.workflow_fit_statistical_model_to_patient import (
     WorkflowFitStatisticalModelToPatient,
 )
 
+# nnUNetv2 (used by TotalSegmentator inside several workflows) spawns a
+# multiprocessing.Pool. On Windows the spawn start method re-imports this
+# script in each child; without the __name__ == "__main__" guard around
+# top-level work, that re-import fires the segmenter again and Python's
+# spawn-cascade detector raises RuntimeError. Wrapping consistently across
+# tutorials also matches the style of tutorial_01.
+if __name__ == "__main__":
+    # %%
+    # Data directory specification
+    REPO_ROOT = Path(__file__).resolve().parent.parent
+    TUTORIALS_DIR = Path(__file__).resolve().parent
+    DATA_DIR = REPO_ROOT / "data"
+    FULL_DATA_DIR = DATA_DIR / "KCL-Heart-Model"
+    TEST_DATA_DIR = DATA_DIR / "test" / "KCL-Heart-Model"
+    OUTPUT_DIR = TUTORIALS_DIR / "output" / "tutorial_04"
+    BASELINES_DIR = REPO_ROOT / "tests" / "baselines"
+    PCA_JSON = TUTORIALS_DIR / "output" / "tutorial_03" / "pca_model.json"
+    LOG_LEVEL = logging.INFO
+
+    # %%
+    # Data reading
+    test_mode = TestTools.running_as_test()
+
+    data_dir = TEST_DATA_DIR if test_mode else FULL_DATA_DIR
+    output_dir = OUTPUT_DIR
+    pca_json = PCA_JSON
+    log_level = LOG_LEVEL
 
-def _extract_surface(mesh: pv.DataSet) -> pv.PolyData:
-    """Extract a PolyData surface using the library's standard VTK algorithm."""
-    if isinstance(mesh, pv.PolyData):
-        return mesh
-    return cast(pv.PolyData, mesh.extract_surface(algorithm="dataset_surface"))
-
-
-def run_tutorial(
-    data_dir: Path,
-    output_dir: Path,
-    *,
-    pca_json: Optional[Path] = None,
-    log_level: int = logging.INFO,
-) -> dict[str, Any]:
-    """Run Tutorial 4: Fit Statistical Shape Model to Patient Data.
-
-    Args:
-        data_dir: Root of the ``data/`` directory (see data/README.md).
-        output_dir: Directory to write outputs and screenshots.
-        pca_json: Optional PCA model JSON produced by Tutorial 3.
-        log_level: Python logging level.
-
-    Returns:
-        dict with keys:
-
-        - ``'registered_model'`` (pv.PolyData): fitted template surface.
-        - ``'registered_file'`` (Path): path to saved ``.vtp``.
-        - ``'screenshots'`` (list[Path]): paths to saved PNG screenshots.
-    """
     output_dir.mkdir(parents=True, exist_ok=True)
 
-    kcl_dir = data_dir / "KCL-Heart-Model"
-    template_file = kcl_dir / "pca_mean.vtu"
+    template_file = data_dir / "pca_mean.vtu"
     if not template_file.exists():
         raise FileNotFoundError(
             f"KCL-Heart-Model template not found: {template_file}\n"
-            "See data/README.md for manual download instructions."
+            "See data/README.md for download instructions."
         )
 
-    template_model = pv.read(str(template_file))
-    template_model = _extract_surface(template_model)
+    template_data = cast(pv.DataSet, pv.read(str(template_file)))
+    if isinstance(template_data, pv.PolyData):
+        template_model = template_data
+    else:
+        template_model = cast(
+            pv.PolyData,
+            template_data.extract_surface(algorithm="dataset_surface"),
+        )
 
-    # Use a subset of sample meshes as stand-in patient models
-    sample_files = sorted((kcl_dir / "sample_meshes").glob("*.vtu"))[:3]
+    sample_files = sorted((data_dir / "sample_meshes").glob("*.vtu"))
     if not sample_files:
-        sample_files = sorted(kcl_dir.glob("*.vtu"))[:3]
+        sample_files = sorted(data_dir.glob("*.vtu"))
+    sample_files = [path for path in sample_files if path.name != "pca_mean.vtu"]
+    sample_files = sample_files[:3]
     if not sample_files:
         raise FileNotFoundError(
-            f"No sample meshes found under {kcl_dir}.\n"
-            "See data/README.md for manual download instructions."
+            f"No patient-like sample meshes found under {data_dir}.\n"
+            "See data/README.md for download instructions."
         )
-    patient_models = []
+
+    patient_models: list[pv.PolyData] = []
     for sample_file in sample_files:
-        sample_model = pv.read(str(sample_file))
-        patient_models.append(_extract_surface(sample_model))
+        sample_data = cast(pv.DataSet, pv.read(str(sample_file)))
+        if isinstance(sample_data, pv.PolyData):
+            patient_models.append(sample_data)
+        else:
+            patient_models.append(
+                cast(
+                    pv.PolyData,
+                    sample_data.extract_surface(algorithm="dataset_surface"),
+                )
+            )
+
+    pca_model: dict[str, Any] | None = None
+    if pca_json.exists():
+        with pca_json.open(encoding="utf-8") as f:
+            pca_model = json.load(f)
 
+    # %%
+    # Workflow initialization
     workflow = WorkflowFitStatisticalModelToPatient(
         template_model=template_model,
         patient_models=patient_models,
         log_level=log_level,
     )
-    if pca_json is not None and pca_json.exists():
-        with open(pca_json, encoding="utf-8") as f:
-            pca_model = json.load(f)
+    if pca_model is not None:
         workflow.set_use_pca_registration(True, pca_model=pca_model)
+
+    # %%
+    # Workflow execution
     result = workflow.run_workflow()
 
+    # %%
+    # Result saving
     registered_surface: pv.PolyData = result["registered_template_model_surface"]
     registered_file = output_dir / "registered_template.vtp"
     registered_surface.save(str(registered_file))
 
-    screenshots: list[Path] = []
-
     patient_combined = (
         pv.merge(patient_models) if len(patient_models) > 1 else patient_models[0]
     )
+    patient_surface = cast(pv.PolyData, patient_combined)
 
-    # Before: template (blue) + patient (red)
     try:
         pv.start_xvfb()
     except Exception:
         pass
+
+    screenshots: list[Path] = []
+
+    before_path = output_dir / "model_before_registration.png"
     plotter = pv.Plotter(off_screen=True, window_size=[800, 600])
-    plotter.add_mesh(
-        template_model,
-        color="dodgerblue",
-        opacity=0.6,
-        label="Template",
-    )
-    plotter.add_mesh(
-        _extract_surface(patient_combined),
-        color="tomato",
-        opacity=0.6,
-        label="Patient",
-    )
+    plotter.add_mesh(template_model, color="dodgerblue", opacity=0.6)
+    plotter.add_mesh(patient_surface, color="tomato", opacity=0.6)
     plotter.camera_position = "iso"
-    before_path = output_dir / "model_before_registration.png"
-    before_path.parent.mkdir(parents=True, exist_ok=True)
     plotter.screenshot(str(before_path))
     plotter.close()
     screenshots.append(before_path)
 
-    # After: registered template (green) + patient (red)
-    plotter2 = pv.Plotter(off_screen=True, window_size=[800, 600])
-    plotter2.add_mesh(
-        registered_surface, color="limegreen", opacity=0.7, label="Registered"
-    )
-    plotter2.add_mesh(
-        _extract_surface(patient_combined),
-        color="tomato",
-        opacity=0.4,
-        label="Patient",
-    )
-    plotter2.camera_position = "iso"
     after_path = output_dir / "model_after_registration.png"
-    plotter2.screenshot(str(after_path))
-    plotter2.close()
+    plotter = pv.Plotter(off_screen=True, window_size=[800, 600])
+    plotter.add_mesh(registered_surface, color="limegreen", opacity=0.7)
+    plotter.add_mesh(patient_surface, color="tomato", opacity=0.4)
+    plotter.camera_position = "iso"
+    plotter.screenshot(str(after_path))
+    plotter.close()
     screenshots.append(after_path)
 
-    return {
+    TestTools(
+        class_name="tutorial_04_fit_statistical_model_to_patient",
+        results_dir=output_dir,
+        baselines_dir=BASELINES_DIR,
+        log_level=log_level,
+    )
+
+    tutorial_results = {
         "registered_model": registered_surface,
         "registered_file": registered_file,
         "screenshots": screenshots,
     }
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser(
-        description=__doc__,
-        formatter_class=argparse.RawDescriptionHelpFormatter,
-    )
-    parser.add_argument(
-        "--data-dir",
-        type=Path,
-        default=Path("data"),
-        help="Root data directory (default: ./data)",
-    )
-    parser.add_argument(
-        "--output-dir",
-        type=Path,
-        default=Path("output") / "tutorial_04",
-        help="Output directory (default: ./output/tutorial_04)",
-    )
-    parser.add_argument(
-        "--pca-json",
-        type=Path,
-        default=Path("output") / "tutorial_03" / "pca_model.json",
-        help="Optional PCA model JSON produced by Tutorial 3",
-    )
-    args = parser.parse_args()
-
-    results = run_tutorial(args.data_dir, args.output_dir, pca_json=args.pca_json)
-    print(f"Registered model: {results['registered_file']}")
-    print(f"Screenshots:      {[str(p) for p in results['screenshots']]}")
diff --git a/tutorials/tutorial_05_vtk_to_usd.py b/tutorials/tutorial_05_vtk_to_usd.py
index 0d5cf45..7676d47 100644
--- a/tutorials/tutorial_05_vtk_to_usd.py
+++ b/tutorials/tutorial_05_vtk_to_usd.py
@@ -1,123 +1,75 @@
 """
-Tutorial 5: VTK Surface Series to Animated USD
+Tutorial 5: VTK Surface Series to USD
 
 Purpose
 -------
-Convert one or more VTK surface files into a USD file suitable for NVIDIA
-Omniverse. Supports a single static mesh, a time-series (animated) set of
-meshes, or a mesh with scalar data visualised via a colormap. This tutorial
-uses the surface files produced by Tutorial 2 (CT Segmentation to VTK) as
-input, but any VTK/VTP/VTU files will work.
-
-Inputs
-------
-- One or more VTK-compatible surface files (``.vtp`` / ``.vtk`` / ``.vtu``).
-  This tutorial looks for ``output/tutorial_02/patient_surfaces.vtp`` (output
-  of Tutorial 2). If that file does not exist, it falls back to any ``.vtp``
-  under ``data/``.
-
-Outputs
--------
-- ``output_dir/surfaces.usd`` - USD file with anatomy materials applied
-- Screenshots (PNG):
-  - ``usd_mesh_rendering.png`` - PyVista off-screen render of the mesh
-
-Strengths
----------
-- Supports time-varying USD for animated sequences (one VTK file per frame).
-- Three appearance modes: ``solid`` (flat colour), ``anatomy`` (clinical
-  material by anatomy type), and ``colormap`` (scalar field visualisation).
-- Coordinate system is automatically converted from RAS to Omniverse Y-up.
-
-Weaknesses / Limitations
-------------------------
-- Requires Tutorial 2 output (or any VTK file) as input; not standalone.
-- Time-series ordering relies on a filename regex pattern (``\.t\d+\.vtp$``);
-  non-conforming filenames are treated as static single-frame input.
-- USD materials use UsdPreviewSurface; advanced Omniverse MDL materials require
-  additional post-processing.
-
-Classes Used
-------------
-- WorkflowConvertVTKToUSD (workflow_convert_vtk_to_usd.py):
-    Loads VTK files, splits meshes, converts to USD, and applies appearance.
-- ConvertVTKToUSD (convert_vtk_to_usd.py):
-    High-level PyVista-to-USD converter with colormap support (used internally).
-- USDAnatomyTools (usd_anatomy_tools.py):
-    Applies clinical material colours to USD prims (used internally).
-
-CLI Equivalent
---------------
-The same main outputs (without screenshots) can be produced via the CLI::
-
-    physiomotion4d-convert-vtk-to-usd \\
-        --input output/tutorial_02/patient_surfaces.vtp \\
-        --output-usd ./output/tutorial_05/surfaces.usd \\
-        --appearance anatomy \\
-        --anatomy-type heart
-
-See ``src/physiomotion4d/cli/convert_vtk_to_usd.py`` for full CLI documentation.
+Convert the VTK surface output from Tutorial 2, or another VTK-compatible mesh,
+into a USD file with anatomy materials.
 
 Data Required
 -------------
-This tutorial uses the output of Tutorial 2 (``output/tutorial_02/patient_surfaces.vtp``).
-Run Tutorial 2 first, or provide any VTK surface file via the ``--vtk-file`` flag.
+Preferred input: ``tutorials/output/tutorial_02/patient_surfaces.vtp``
+Fallback input: any ``*.vtp`` under ``data`` or ``data/test``
 """
 
+# %%
+# Imports
 from __future__ import annotations
 
-import argparse
 import logging
 from pathlib import Path
-from typing import Any, Optional
-
-import pyvista as pv
+from typing import Optional
 
 from physiomotion4d.test_tools import TestTools
 from physiomotion4d.workflow_convert_vtk_to_usd import WorkflowConvertVTKToUSD
 
+# nnUNetv2 (used by TotalSegmentator inside several workflows) spawns a
+# multiprocessing.Pool. On Windows the spawn start method re-imports this
+# script in each child; without the __name__ == "__main__" guard around
+# top-level work, that re-import fires the segmenter again and Python's
+# spawn-cascade detector raises RuntimeError. Wrapping consistently across
+# tutorials also matches the style of tutorial_01.
+if __name__ == "__main__":
+    # %%
+    # Data directory specification
+    REPO_ROOT = Path(__file__).resolve().parent.parent
+    TUTORIALS_DIR = Path(__file__).resolve().parent
+    DATA_DIR = REPO_ROOT / "data"
+    FULL_DATA_DIR = DATA_DIR
+    TEST_DATA_DIR = DATA_DIR / "test"
+    OUTPUT_DIR = TUTORIALS_DIR / "output" / "tutorial_05"
+    BASELINES_DIR = REPO_ROOT / "tests" / "baselines"
+    TUTORIAL_02_SURFACE = (
+        TUTORIALS_DIR / "output" / "tutorial_02" / "patient_surfaces.vtp"
+    )
+    VTK_FILE: Optional[Path] = None
+    LOG_LEVEL = logging.INFO
 
-def run_tutorial(
-    data_dir: Path,
-    output_dir: Path,
-    *,
-    vtk_file: Optional[Path] = None,
-    log_level: int = logging.INFO,
-) -> dict[str, Any]:
-    """Run Tutorial 5: VTK Surface Series to Animated USD.
-
-    Args:
-        data_dir: Root of the ``data/`` directory (see data/README.md).
-        output_dir: Directory to write outputs and screenshots.
-        vtk_file: Explicit path to a VTK file; overrides auto-discovery.
-        log_level: Python logging level.
+    # %%
+    # Data reading
+    test_mode = TestTools.running_as_test()
 
-    Returns:
-        dict with keys:
+    data_dir = TEST_DATA_DIR if test_mode else FULL_DATA_DIR
+    output_dir = OUTPUT_DIR
+    vtk_file = VTK_FILE
+    log_level = LOG_LEVEL
 
-        - ``'usd_file'`` (str): path to the output USD file.
-        - ``'screenshots'`` (list[Path]): paths to saved PNG screenshots.
-    """
     output_dir.mkdir(parents=True, exist_ok=True)
 
+    if vtk_file is None and TUTORIAL_02_SURFACE.exists():
+        vtk_file = TUTORIAL_02_SURFACE
     if vtk_file is None:
-        # Prefer Tutorial 2 output
-        project_root = data_dir.parent
-        candidate = project_root / "output" / "tutorial_02" / "patient_surfaces.vtp"
-        if candidate.exists():
-            vtk_file = candidate
-        else:
-            # Fall back to any .vtp under data/
-            found = list(data_dir.rglob("*.vtp"))
-            if not found:
-                raise FileNotFoundError(
-                    "No VTK file found. Run Tutorial 2 first, or specify "
-                    "--vtk-file <path>."
-                )
-            vtk_file = found[0]
-
+        vtk_candidates = sorted(data_dir.rglob("*.vtp"))
+        if not vtk_candidates:
+            raise FileNotFoundError(
+                "No VTK surface file found. Run Tutorial 2 first or place a "
+                f"*.vtp file under {data_dir}."
+            )
+        vtk_file = vtk_candidates[0]
+
+    # %%
+    # Workflow initialization
     output_usd = output_dir / "surfaces.usd"
-
     workflow = WorkflowConvertVTKToUSD(
         vtk_files=[vtk_file],
         output_usd=output_usd,
@@ -126,58 +78,26 @@ def run_tutorial(
         separate_by_connectivity=True,
         log_level=log_level,
     )
-    usd_path = workflow.run()
 
-    # Screenshots
+    # %%
+    # Workflow execution
+    usd_file = workflow.run()
+
+    # %%
+    # Result saving
     tt = TestTools(
+        class_name="tutorial_05_vtk_to_usd",
         results_dir=output_dir,
-        baselines_dir=output_dir / "baselines",
-        class_name="tutorial_05",
-        results_output_dir=output_dir,
+        baselines_dir=BASELINES_DIR,
         log_level=log_level,
     )
 
     screenshots: list[Path] = []
-
-    mesh = pv.read(str(vtk_file))
     screenshots.append(
-        tt.save_screenshot_mesh(
-            mesh,
+        tt.save_screenshot_openusd(
+            usd_file,
             "usd_mesh_rendering.png",
-            camera_position="iso",
-            color="lightcoral",
-            opacity=0.9,
         )
     )
 
-    return {"usd_file": usd_path, "screenshots": screenshots}
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser(
-        description=__doc__,
-        formatter_class=argparse.RawDescriptionHelpFormatter,
-    )
-    parser.add_argument(
-        "--data-dir",
-        type=Path,
-        default=Path("data"),
-        help="Root data directory (default: ./data)",
-    )
-    parser.add_argument(
-        "--output-dir",
-        type=Path,
-        default=Path("output") / "tutorial_05",
-        help="Output directory (default: ./output/tutorial_05)",
-    )
-    parser.add_argument(
-        "--vtk-file",
-        type=Path,
-        default=None,
-        help="Explicit VTK input file (default: auto-discover from Tutorial 2 output)",
-    )
-    args = parser.parse_args()
-
-    results = run_tutorial(args.data_dir, args.output_dir, vtk_file=args.vtk_file)
-    print(f"USD file:    {results['usd_file']}")
-    print(f"Screenshots: {[str(p) for p in results['screenshots']]}")
+    tutorial_results = {"usd_file": usd_file, "screenshots": screenshots}
diff --git a/tutorials/tutorial_06_reconstruct_highres_4d_ct.py b/tutorials/tutorial_06_reconstruct_highres_4d_ct.py
index 586c882..1fcbe9c 100644
--- a/tutorials/tutorial_06_reconstruct_highres_4d_ct.py
+++ b/tutorials/tutorial_06_reconstruct_highres_4d_ct.py
@@ -3,88 +3,23 @@
 
 Purpose
 -------
-Reconstruct a high-resolution dynamic 4D CT volume from a time series of
-lower-resolution or sparse CT frames. The workflow registers each time frame
-to a high-resolution reference image, producing a sequence of reconstructed
-volumes that share the spatial resolution of the reference. This is useful for
-respiratory-gated lung CT (DirLab-4DCT) where breath-hold reference scans are
-available alongside lower-quality respiratory-phase images.
-
-Inputs
-------
-- A list of 3D CT images (``itk.Image``): the time series to reconstruct.
-  Expected location: ``data/DirLab-4DCT/Case1/`` (T00-T90 phases).
-- A high-resolution fixed reference image (``itk.Image``):
-  the target space for reconstruction.
-  Expected location: ``data/DirLab-4DCT/Case1/`` (any phase used as reference).
-
-Note
-----
-The DirLab-4DCT sample data used by this tutorial does not include a separate
-high-resolution breath-hold reference image. For demonstration and regression
-testing, the tutorial uses one available DirLab respiratory phase as the fixed
-reference, so the reconstructed outputs inherit that phase image's resolution
-rather than true higher-resolution reference spacing.
-
-Outputs
--------
-- ``output_dir/reconstructed_frame_<N>.mha`` - one reconstructed 3D image per frame
-- Screenshots (PNG):
-  - ``reference_frame.png`` - axial slice of the high-resolution reference image
-  - ``reconstructed_frame.png`` - axial slice of the first reconstructed frame
-
-Strengths
----------
-- Bidirectional propagation of registration from the reference frame reduces
-  accumulated error for frames far from the reference.
-- Temporal smoothing via ``prior_weight`` parameter reduces frame-to-frame jitter.
-- Supports ``'ants'``, ``'icon'``, and ``'ants_icon'`` (two-stage) registration.
-
-Weaknesses / Limitations
-------------------------
-- Requires the DirLab-4DCT dataset (manual download; see data/README.md).
-- ICON registration (default part of ``'ants_icon'``) requires a GPU.
-- Reconstruction quality is bounded by the accuracy of the registration; large
-  respiratory excursion between phases can cause residual artefacts.
-- Runtime is proportional to the number of frames times registration cost.
-
-Classes Used
-------------
-- WorkflowReconstructHighres4DCT (workflow_reconstruct_highres_4d_ct.py):
-    Registers each time frame to the fixed reference and reconstructs volumes.
-- RegisterTimeSeriesImages (register_time_series_images.py):
-    Chains frame-to-frame registration with optional temporal smoothing (used
-    internally).
-- RegisterImagesANTs / RegisterImagesICON (register_images_ants.py / _icon.py):
-    Individual frame registration backends (used internally).
-
-CLI Equivalent
---------------
-The same main outputs (without screenshots) can be produced via the CLI::
-
-    physiomotion4d-reconstruct-highres-4d-ct \\
-        --time-series-dir data/DirLab-4DCT/Case1 \\
-        --fixed-image data/DirLab-4DCT/Case1/case1_T00.mhd \\
-        --registration-method ants \\
-        --output-dir ./output/tutorial_06
-
-See ``src/physiomotion4d/cli/reconstruct_highres_4d_ct.py`` for full CLI
-documentation.
+Register a short CT time series to a fixed reference image and save the
+reconstructed frames. DirLab does not provide a separate high-resolution
+breath-hold reference image, so this tutorial uses one available respiratory
+phase as the fixed reference.
 
 Data Required
 -------------
-See data/README.md for download instructions and dataset licensing.
-Dataset: DirLab 4D-CT - https://www.dir-lab.com/ReferenceData.html
-Manual download required. Place files under ``data/DirLab-4DCT/`` as described
-in data/README.md.
+Full data: ``data/DirLab-4DCT/Case1``
+Test data: ``data/test/DirLab-4DCT/Case1``
 """
 
+# %%
+# Imports
 from __future__ import annotations
 
-import argparse
 import logging
 from pathlib import Path
-from typing import Any
 
 import itk
 
@@ -93,57 +28,57 @@
     WorkflowReconstructHighres4DCT,
 )
 
+# nnUNetv2 (used by TotalSegmentator inside several workflows) spawns a
+# multiprocessing.Pool. On Windows the spawn start method re-imports this
+# script in each child; without the __name__ == "__main__" guard around
+# top-level work, that re-import fires the segmenter again and Python's
+# spawn-cascade detector raises RuntimeError. Wrapping consistently across
+# tutorials also matches the style of tutorial_01.
+if __name__ == "__main__":
+    # %%
+    # Data directory specification
+    REPO_ROOT = Path(__file__).resolve().parent.parent
+    TUTORIALS_DIR = Path(__file__).resolve().parent
+    DATA_DIR = REPO_ROOT / "data"
+    FULL_DATA_DIR = DATA_DIR / "DirLab-4DCT" / "Case1"
+    TEST_DATA_DIR = DATA_DIR / "test" / "DirLab-4DCT" / "Case1"
+    OUTPUT_DIR = TUTORIALS_DIR / "output" / "tutorial_06"
+    BASELINES_DIR = REPO_ROOT / "tests" / "baselines"
+    MAX_FRAMES = 4
+    REGISTRATION_METHOD = "ants"
+    LOG_LEVEL = logging.INFO
+
+    # %%
+    # Data reading
+    test_mode = TestTools.running_as_test()
+
+    data_dir = TEST_DATA_DIR if test_mode else FULL_DATA_DIR
+    output_dir = OUTPUT_DIR
+    registration_method = REGISTRATION_METHOD
+    log_level = LOG_LEVEL
+
+    if test_mode:
+        max_frames = min(MAX_FRAMES, 3)
+        number_of_iterations_ants = [1, 0]
+    else:
+        max_frames = MAX_FRAMES
+        number_of_iterations_ants = [30, 15, 7, 3]
 
-def run_tutorial(
-    data_dir: Path,
-    output_dir: Path,
-    *,
-    case: int = 1,
-    max_frames: int = 4,
-    registration_method: str = "ants",
-    log_level: int = logging.INFO,
-) -> dict[str, Any]:
-    """Run Tutorial 6: Reconstruct High-Resolution 4D CT.
-
-    Args:
-        data_dir: Root of the ``data/`` directory (see data/README.md).
-        output_dir: Directory to write outputs and screenshots.
-        case: DirLab case number (1-10). Default: 1.
-        max_frames: Maximum number of time frames to reconstruct (for speed).
-        registration_method: ``'ants'`` (CPU-capable) or ``'icon'`` (GPU) or
-            ``'ants_icon'`` (two-stage). Default: ``'ants'``.
-        log_level: Python logging level.
-
-    Returns:
-        dict with keys:
-
-        - ``'reconstructed_images'`` (list[itk.Image]): reconstructed volumes.
-        - ``'reconstructed_files'`` (list[Path]): saved ``.mha`` paths.
-        - ``'screenshots'`` (list[Path]): paths to saved PNG screenshots.
-    """
     output_dir.mkdir(parents=True, exist_ok=True)
 
-    dirlab_dir = data_dir / "DirLab-4DCT"
-    case_dir = dirlab_dir / f"Case{case}"
-
-    # Discover phase images (MetaImage .mhd or .mha)
-    phase_files = sorted(list(case_dir.glob("*.mhd")) + list(case_dir.glob("*.mha")))
-    if not phase_files:
-        phase_files = sorted(
-            list(dirlab_dir.glob(f"Case{case}Pack_T*.mhd"))
-            + list(dirlab_dir.glob(f"Case{case}Pack_T*.mha"))
-        )
+    phase_files = sorted(list(data_dir.glob("*.mhd")) + list(data_dir.glob("*.mha")))
     if not phase_files:
         raise FileNotFoundError(
-            f"No .mhd / .mha files found for DirLab-4DCT case {case} under "
-            f"{case_dir} or {dirlab_dir}.\n"
-            "See data/README.md for manual download instructions."
+            f"No DirLab phase images found under {data_dir}.\n"
+            "See data/README.md for download instructions."
         )
 
     phase_files = phase_files[:max_frames]
-    time_series = [itk.imread(str(f)) for f in phase_files]
-    fixed_image = time_series[0]  # use first phase as high-res reference
+    time_series = [itk.imread(str(path)) for path in phase_files]
+    fixed_image = time_series[0]
 
+    # %%
+    # Workflow initialization
     workflow = WorkflowReconstructHighres4DCT(
         time_series_images=time_series,
         fixed_image=fixed_image,
@@ -152,26 +87,29 @@ def run_tutorial(
         log_level=log_level,
     )
     workflow.set_modality("ct")
+    workflow.set_number_of_iterations_ants(number_of_iterations_ants)
+
+    # %%
+    # Workflow execution
     result = workflow.run_workflow()
 
-    reconstructed: list[itk.Image] = result["reconstructed_images"]
+    # %%
+    # Result saving
+    reconstructed_images: list[itk.Image] = result["reconstructed_images"]
     reconstructed_files: list[Path] = []
-    for i, vol in enumerate(reconstructed):
-        out_path = output_dir / f"reconstructed_frame_{i:03d}.mha"
-        itk.imwrite(vol, str(out_path), compression=True)
+    for frame_index, image in enumerate(reconstructed_images):
+        out_path = output_dir / f"reconstructed_frame_{frame_index:03d}.mha"
+        itk.imwrite(image, str(out_path), compression=True)
         reconstructed_files.append(out_path)
 
-    # Screenshots
     tt = TestTools(
+        class_name="tutorial_06_reconstruct_highres_4d_ct",
         results_dir=output_dir,
-        baselines_dir=output_dir / "baselines",
-        class_name="tutorial_06",
-        results_output_dir=output_dir,
+        baselines_dir=BASELINES_DIR,
         log_level=log_level,
     )
 
     screenshots: list[Path] = []
-
     screenshots.append(
         tt.save_screenshot_image_slice(
             fixed_image,
@@ -181,11 +119,10 @@ def run_tutorial(
             colormap="gray",
         )
     )
-
-    if reconstructed:
+    if reconstructed_images:
         screenshots.append(
             tt.save_screenshot_image_slice(
-                reconstructed[0],
+                reconstructed_images[0],
                 "reconstructed_frame.png",
                 axis=0,
                 slice_fraction=0.5,
@@ -193,50 +130,8 @@ def run_tutorial(
             )
         )
 
-    return {
-        "reconstructed_images": reconstructed,
+    tutorial_results = {
+        "reconstructed_images": reconstructed_images,
         "reconstructed_files": reconstructed_files,
         "screenshots": screenshots,
     }
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser(
-        description=__doc__,
-        formatter_class=argparse.RawDescriptionHelpFormatter,
-    )
-    parser.add_argument(
-        "--data-dir",
-        type=Path,
-        default=Path("data"),
-        help="Root data directory (default: ./data)",
-    )
-    parser.add_argument(
-        "--output-dir",
-        type=Path,
-        default=Path("output") / "tutorial_06",
-        help="Output directory (default: ./output/tutorial_06)",
-    )
-    parser.add_argument(
-        "--case",
-        type=int,
-        default=1,
-        choices=list(range(1, 11)),
-        help="DirLab case number 1-10 (default: 1)",
-    )
-    parser.add_argument(
-        "--registration-method",
-        default="ants",
-        choices=["ants", "icon", "ants_icon"],
-        help="Registration method (default: ants)",
-    )
-    args = parser.parse_args()
-
-    results = run_tutorial(
-        args.data_dir,
-        args.output_dir,
-        case=args.case,
-        registration_method=args.registration_method,
-    )
-    print(f"Reconstructed frames: {[str(p) for p in results['reconstructed_files']]}")
-    print(f"Screenshots:          {[str(p) for p in results['screenshots']]}")
diff --git a/tutorials/tutorial_07_dirlab_pca_model.py b/tutorials/tutorial_07_dirlab_pca_model.py
new file mode 100644
index 0000000..edfae3d
--- /dev/null
+++ b/tutorials/tutorial_07_dirlab_pca_model.py
@@ -0,0 +1,305 @@
+"""Tutorial 7: Build and fit a PCA lung-lobe model from DirLab 4D CT cases.
+
+This tutorial uses one respiratory phase from each available DirLab case. It
+segments the five lung lobes, builds a surface PCA model from all but two
+cases, then fits that model to every available case.
+"""
+
+# %%
+from __future__ import annotations
+
+import json
+import logging
+from pathlib import Path
+from typing import Any, cast
+
+import itk
+import numpy as np
+import pyvista as pv
+
+from physiomotion4d.contour_tools import ContourTools
+from physiomotion4d.workflow_convert_ct_to_vtk import WorkflowConvertCTToVTK
+from physiomotion4d.workflow_create_statistical_model import (
+    WorkflowCreateStatisticalModel,
+)
+from physiomotion4d.test_tools import TestTools
+from physiomotion4d.workflow_fit_statistical_model_to_patient import (
+    WorkflowFitStatisticalModelToPatient,
+)
+
+
+# nnUNetv2 (used by TotalSegmentator inside several workflows) spawns a
+# multiprocessing.Pool. On Windows the spawn start method re-imports this
+# script in each child; without the __name__ == "__main__" guard around
+# top-level work, that re-import fires the segmenter again and Python's
+# spawn-cascade detector raises RuntimeError. Wrapping consistently across
+# tutorials also matches the style of tutorial_01.
+if __name__ == "__main__":
+    # %%
+    REPO_ROOT = Path(__file__).resolve().parent.parent
+    TUTORIALS_DIR = Path(__file__).resolve().parent
+    DATA_DIR = REPO_ROOT / "data"
+    FULL_DATA_DIR = DATA_DIR
+    TEST_DATA_DIR = DATA_DIR / "test"
+    OUTPUT_DIR = TUTORIALS_DIR / "output" / "tutorial_07_dirlab_pca_model"
+    PCA_COMPONENTS = 5
+    LOG_LEVEL = logging.INFO
+
+    # %%
+    def create_meshes(
+        data_dir: Path,
+        output_dir: Path,
+        log_level: int = logging.INFO,
+    ) -> dict[str, Path]:
+        """Segment DirLab CT cases and save one five-lobe surface mesh per case.
+
+        Parameters
+        ----------
+        data_dir
+            Repository data directory containing ``DirLab-4DCT``.
+        output_dir
+            Directory where tutorial outputs are written.
+        log_level
+            Logging level for the conversion workflow.
+
+        Returns
+        -------
+        dict[str, Path]
+            Output directory and saved lung-lobe surface mesh filenames.
+        """
+
+        dirlab_dir = data_dir / "DirLab-4DCT"
+        meshes_dir = output_dir / "meshes"
+        meshes_dir.mkdir(parents=True, exist_ok=True)
+
+        case_prefixes = [
+            "Case1Pack",
+            "Case2Pack",
+            "Case3Pack",
+            "Case4Pack",
+            "Case5Pack",
+            "Case6Pack",
+            "Case7Pack",
+            "Case8Deploy",
+            "Case9Pack",
+            "Case10Pack",
+        ]
+        lung_lobe_ids = {
+            10: "lung_upper_lobe_left",
+            11: "lung_lower_lobe_left",
+            12: "lung_upper_lobe_right",
+            13: "lung_middle_lobe_right",
+            14: "lung_lower_lobe_right",
+        }
+
+        mesh_files: list[Path] = []
+        contour_tools = ContourTools(log_level=log_level)
+        for case_number, case_prefix in enumerate(case_prefixes, start=1):
+            mesh_file = meshes_dir / f"{case_prefix}_lung_lobes.vtp"
+            if mesh_file.exists():
+                mesh_files.append(mesh_file)
+                continue
+
+            case_dir = dirlab_dir / f"Case{case_number}"
+            phase_files = sorted(case_dir.glob("*.mha")) + sorted(
+                case_dir.glob("*.mhd")
+            )
+            if not phase_files:
+                phase_files = sorted(dirlab_dir.glob(f"{case_prefix}_T*.mha"))
+                phase_files += sorted(dirlab_dir.glob(f"{case_prefix}_T*.mhd"))
+            if not phase_files and case_number == 8:
+                phase_files = sorted(dirlab_dir.glob("Case8Pack_T*.mha"))
+                phase_files += sorted(dirlab_dir.glob("Case8Pack_T*.mhd"))
+            if not phase_files:
+                print(f"Skipping {case_prefix}: no DirLab phase image found")
+                continue
+
+            print(f"Segmenting {case_prefix} from {phase_files[0].name}")
+            image = itk.imread(str(phase_files[0]))
+            workflow = WorkflowConvertCTToVTK(
+                segmentation_method="total_segmentator",
+                log_level=log_level,
+            )
+            result = workflow.run_workflow(
+                input_image=image,
+                contrast_enhanced_study=False,
+                anatomy_groups=["lung"],
+            )
+
+            labelmap = result["labelmap"]
+            labelmap_arr = itk.GetArrayFromImage(labelmap)
+            lobe_surfaces: list[pv.PolyData] = []
+            for label_id, lobe_name in lung_lobe_ids.items():
+                lobe_arr = (labelmap_arr == label_id).astype(np.uint8)
+                if int(lobe_arr.sum()) == 0:
+                    print(f"Skipping {case_prefix}: missing {lobe_name}")
+                    lobe_surfaces = []
+                    break
+
+                lobe_mask = itk.GetImageFromArray(lobe_arr)
+                lobe_mask.CopyInformation(labelmap)
+                lobe_surface = contour_tools.extract_contours(lobe_mask)
+                lobe_surface.field_data["LungLobeName"] = np.array([lobe_name])
+                lobe_surface.field_data["LungLobeLabel"] = np.array([label_id])
+                lobe_surfaces.append(lobe_surface)
+
+            if len(lobe_surfaces) != len(lung_lobe_ids):
+                continue
+
+            lung_surface = cast(
+                pv.PolyData,
+                pv.merge(lobe_surfaces, merge_points=False),
+            )
+            lung_surface.save(mesh_file)
+            mesh_files.append(mesh_file)
+
+        return {"meshes_dir": meshes_dir, **{mesh.stem: mesh for mesh in mesh_files}}
+
+    def create_model(
+        meshes_dir: Path,
+        output_dir: Path,
+        pca_components: int = 5,
+        log_level: int = logging.INFO,
+    ) -> dict[str, Any]:
+        """Create a surface PCA model from all but the final two DirLab lobe meshes.
+
+        Parameters
+        ----------
+        meshes_dir
+            Directory containing lung-lobe surface files from :func:`create_meshes`.
+        output_dir
+            Directory where model outputs are written.
+        pca_components
+            Requested PCA component count.
+        log_level
+            Logging level for the model creation workflow.
+
+        Returns
+        -------
+        dict[str, Any]
+            PCA model, saved model filenames, training files, and all mesh filenames.
+        """
+
+        model_dir = output_dir / "pca_model"
+        model_dir.mkdir(parents=True, exist_ok=True)
+
+        mesh_files = sorted(meshes_dir.glob("*_lung_lobes.vtp"))
+        if len(mesh_files) < 4:
+            raise ValueError("At least four DirLab lung-lobe meshes are needed.")
+
+        training_files = mesh_files[:-2]
+        held_out_files = mesh_files[-2:]
+        sample_meshes = [pv.read(str(mesh_file)) for mesh_file in training_files]
+        reference_mesh = pv.read(str(training_files[0]))
+        component_count = min(pca_components, max(1, len(sample_meshes) - 1))
+
+        workflow = WorkflowCreateStatisticalModel(
+            sample_meshes=sample_meshes,
+            reference_mesh=reference_mesh,
+            pca_number_of_components=component_count,
+            solve_for_surface_pca=True,
+            log_level=log_level,
+        )
+        result = workflow.run_workflow()
+
+        mean_surface_file = model_dir / "pca_mean_surface.vtp"
+        pca_model_file = model_dir / "pca_model.json"
+        result["pca_mean_surface"].save(mean_surface_file)
+        pca_model_file.write_text(
+            json.dumps(result["pca_model"], indent=2),
+            encoding="utf-8",
+        )
+
+        return {
+            "pca_model": result["pca_model"],
+            "mean_surface_file": mean_surface_file,
+            "pca_model_file": pca_model_file,
+            "training_files": training_files,
+            "held_out_files": held_out_files,
+            "mesh_files": mesh_files,
+        }
+
+    def fit_model(
+        model_result: dict[str, Any],
+        output_dir: Path,
+        log_level: int = logging.INFO,
+    ) -> dict[str, Path]:
+        """Fit the surface PCA lung-lobe model to every available DirLab mesh.
+
+        Parameters
+        ----------
+        model_result
+            Dictionary returned by :func:`create_model`.
+        output_dir
+            Directory where fit outputs are written.
+        log_level
+            Logging level for the fitting workflow.
+
+        Returns
+        -------
+        dict[str, Path]
+            Saved fitted surface filenames.
+        """
+
+        fits_dir = output_dir / "fits"
+        fits_dir.mkdir(parents=True, exist_ok=True)
+
+        template_model = pv.read(str(model_result["mean_surface_file"]))
+        fitted_files: dict[str, Path] = {}
+
+        for patient_file in model_result["mesh_files"]:
+            patient_model = pv.read(str(patient_file))
+            workflow = WorkflowFitStatisticalModelToPatient(
+                template_model=template_model,
+                patient_models=[patient_model],
+                log_level=log_level,
+            )
+            workflow.set_use_pca_registration(
+                True,
+                pca_model=model_result["pca_model"],
+                pca_number_of_modes=0,
+                pca_uses_surface=True,
+            )
+            workflow.set_use_mask_to_mask_registration(False)
+
+            result = workflow.run_workflow()
+            fitted_surface = result["registered_template_model_surface"]
+            fitted_file = fits_dir / f"{patient_file.stem}_pca_fit.vtp"
+            fitted_surface.save(fitted_file)
+            fitted_files[patient_file.stem] = fitted_file
+
+        return fitted_files
+
+    def run_tutorial() -> dict[str, Any]:
+        """Run mesh creation, PCA model creation, and PCA fitting in sequence.
+
+        Returns
+        -------
+        dict[str, Any]
+            Mesh, PCA model, and fitted model output information.
+        """
+
+        data_dir = TEST_DATA_DIR if TestTools.running_as_test() else FULL_DATA_DIR
+        output_dir = OUTPUT_DIR
+        pca_components = PCA_COMPONENTS
+        log_level = LOG_LEVEL
+
+        output_dir.mkdir(parents=True, exist_ok=True)
+        mesh_result = create_meshes(data_dir, output_dir, log_level=log_level)
+        model_result = create_model(
+            mesh_result["meshes_dir"],
+            output_dir,
+            pca_components=pca_components,
+            log_level=log_level,
+        )
+        fit_result = fit_model(model_result, output_dir, log_level=log_level)
+
+        return {
+            "mesh_result": mesh_result,
+            "model_result": model_result,
+            "fit_result": fit_result,
+        }
+
+    # %%
+    # Run this cell in VS Code or Cursor:
+    tutorial_results = run_tutorial()
diff --git a/tutorials/tutorial_08_dirlab_pca_time_series.py b/tutorials/tutorial_08_dirlab_pca_time_series.py
new file mode 100644
index 0000000..3a544f2
--- /dev/null
+++ b/tutorials/tutorial_08_dirlab_pca_time_series.py
@@ -0,0 +1,199 @@
+"""
+Tutorial 8: Propagate DirLab PCA lung-lobe fits through each 4D CT time series.
+
+This tutorial uses the per-case PCA-fitted reference meshes created by Tutorial
+7. For each DirLab case, it registers every respiratory phase to the reference
+phase used by Tutorial 7, saves the image transforms, and applies the
+reference-to-phase transform to the fitted mesh so each time point has a
+PCA-derived lung-lobe surface.
+
+Data Required
+-------------
+See data/README.md for DirLab-4DCT download instructions. Run Tutorial 7 first
+so ``output/tutorial_07_dirlab_pca_model/fits`` contains the per-case fitted
+reference-stage meshes.
+"""
+
+# %%
+from __future__ import annotations
+
+import logging
+from pathlib import Path
+from typing import Any, Optional
+
+import itk
+import pyvista as pv
+
+from physiomotion4d.register_time_series_images import RegisterTimeSeriesImages
+from physiomotion4d.test_tools import TestTools
+from physiomotion4d.transform_tools import TransformTools
+
+
+# nnUNetv2 (used by TotalSegmentator inside several workflows) spawns a
+# multiprocessing.Pool. On Windows the spawn start method re-imports this
+# script in each child; without the __name__ == "__main__" guard around
+# top-level work, that re-import fires the segmenter again and Python's
+# spawn-cascade detector raises RuntimeError. Wrapping consistently across
+# tutorials also matches the style of tutorial_01.
+if __name__ == "__main__":
+    # %%
+    REPO_ROOT = Path(__file__).resolve().parent.parent
+    TUTORIALS_DIR = Path(__file__).resolve().parent
+    DATA_DIR = REPO_ROOT / "data"
+    FULL_DATA_DIR = DATA_DIR
+    TEST_DATA_DIR = DATA_DIR / "test"
+    OUTPUT_DIR = TUTORIALS_DIR / "output" / "tutorial_08_dirlab_pca_time_series"
+    TUTORIAL_07_OUTPUT_DIR = TUTORIALS_DIR / "output" / "tutorial_07_dirlab_pca_model"
+    CASE: Optional[int] = None
+    LOG_LEVEL = logging.INFO
+
+    DIRLAB_CASE_PREFIXES = [
+        "Case1Pack",
+        "Case2Pack",
+        "Case3Pack",
+        "Case4Pack",
+        "Case5Pack",
+        "Case6Pack",
+        "Case7Pack",
+        "Case8Deploy",
+        "Case9Pack",
+        "Case10Pack",
+    ]
+
+    def run_tutorial() -> dict[str, Any]:
+        """Run Tutorial 8: propagate Tutorial 7 PCA fits through DirLab time series.
+
+        Returns
+        -------
+        dict[str, Any]
+            Per-case transform filenames, propagated mesh filenames, and losses.
+        """
+
+        data_dir = TEST_DATA_DIR if TestTools.running_as_test() else FULL_DATA_DIR
+        output_dir = OUTPUT_DIR
+        tutorial_07_output_dir = TUTORIAL_07_OUTPUT_DIR
+        case = CASE
+        log_level = LOG_LEVEL
+
+        output_dir.mkdir(parents=True, exist_ok=True)
+        dirlab_dir = data_dir / "DirLab-4DCT"
+        fits_dir = tutorial_07_output_dir / "fits"
+        transform_tools = TransformTools(log_level=log_level)
+
+        if case is None:
+            case_numbers = list(range(1, 11))
+        else:
+            case_numbers = [case]
+
+        tutorial_outputs: dict[str, Any] = {}
+        for case_number in case_numbers:
+            case_prefix = DIRLAB_CASE_PREFIXES[case_number - 1]
+            case_dir = dirlab_dir / f"Case{case_number}"
+            phase_files = sorted(case_dir.glob("*.mha")) + sorted(
+                case_dir.glob("*.mhd")
+            )
+            if not phase_files:
+                phase_files = sorted(dirlab_dir.glob(f"{case_prefix}_T*.mha"))
+                phase_files += sorted(dirlab_dir.glob(f"{case_prefix}_T*.mhd"))
+            if not phase_files and case_number == 8:
+                phase_files = sorted(dirlab_dir.glob("Case8Pack_T*.mha"))
+                phase_files += sorted(dirlab_dir.glob("Case8Pack_T*.mhd"))
+            if not phase_files:
+                print(f"Skipping {case_prefix}: no DirLab phase images found")
+                continue
+
+            fitted_mesh_file = fits_dir / f"{case_prefix}_lung_lobes_pca_fit.vtp"
+            if not fitted_mesh_file.exists():
+                raise FileNotFoundError(
+                    f"Missing Tutorial 7 fitted mesh: {fitted_mesh_file}. "
+                    "Run Tutorial 7 before Tutorial 8."
+                )
+
+            print(f"Registering {case_prefix}: {len(phase_files)} phases")
+            case_output_dir = output_dir / case_prefix
+            transforms_dir = case_output_dir / "transforms"
+            meshes_dir = case_output_dir / "meshes"
+            transforms_dir.mkdir(parents=True, exist_ok=True)
+            meshes_dir.mkdir(parents=True, exist_ok=True)
+
+            time_series = [itk.imread(str(phase_file)) for phase_file in phase_files]
+            fixed_image = time_series[0]
+
+            registrar = RegisterTimeSeriesImages(
+                registration_method="ants_icon",
+                log_level=log_level,
+            )
+            registrar.set_modality("ct")
+            registrar.set_fixed_image(fixed_image)
+            registration_result = registrar.register_time_series(
+                moving_images=time_series,
+                reference_frame=0,
+                register_reference=False,
+            )
+
+            fitted_reference_mesh = pv.read(str(fitted_mesh_file))
+            case_transform_files: list[Path] = []
+            case_mesh_files: list[Path] = []
+
+            forward_transforms = registration_result["forward_transforms"]
+            inverse_transforms = registration_result["inverse_transforms"]
+            if not (
+                len(phase_files) == len(forward_transforms) == len(inverse_transforms)
+            ):
+                raise ValueError(
+                    f"{case_prefix}: length mismatch between phase_files "
+                    f"({len(phase_files)}), forward_transforms "
+                    f"({len(forward_transforms)}), and inverse_transforms "
+                    f"({len(inverse_transforms)})."
+                )
+
+            for phase_file, phase_to_reference, reference_to_phase in zip(
+                phase_files,
+                forward_transforms,
+                inverse_transforms,
+            ):
+                phase_name = phase_file.stem
+                phase_to_reference_file = transforms_dir / (
+                    f"{phase_name}_phase_to_reference.hdf"
+                )
+                reference_to_phase_file = transforms_dir / (
+                    f"{phase_name}_reference_to_phase.hdf"
+                )
+                itk.transformwrite(
+                    phase_to_reference,
+                    str(phase_to_reference_file),
+                    compression=True,
+                )
+                itk.transformwrite(
+                    reference_to_phase,
+                    str(reference_to_phase_file),
+                    compression=True,
+                )
+
+                phase_mesh = transform_tools.transform_pvcontour(
+                    fitted_reference_mesh,
+                    reference_to_phase,
+                    with_deformation_magnitude=True,
+                )
+                phase_mesh_file = meshes_dir / f"{phase_name}_pca_fit.vtp"
+                phase_mesh.save(phase_mesh_file)
+
+                case_transform_files.extend(
+                    [phase_to_reference_file, reference_to_phase_file]
+                )
+                case_mesh_files.append(phase_mesh_file)
+
+            tutorial_outputs[case_prefix] = {
+                "reference_phase": phase_files[0],
+                "phase_files": phase_files,
+                "fitted_reference_mesh": fitted_mesh_file,
+                "transform_files": case_transform_files,
+                "mesh_files": case_mesh_files,
+                "losses": registration_result["losses"],
+            }
+
+        return tutorial_outputs
+
+    # %%
+    # Run this cell in VS Code or Cursor:
+    tutorial_results = run_tutorial()