Towards Rapid Differentiation of Liver Malignancies using Optical Coherence Tomography and Deep Learning

This repository provides the official PyTorch implementation for the paper "Towards Rapid Differentiation of Liver Malignancies using Optical Coherence Tomography and Deep Learning".

The code trains and evaluates deep learning models (ResNet50 and Xception) to classify Optical Coherence Tomography (OCT) images of liver tissue, differentiating between liver parenchyma and various malignancies like iCCA, CRLM, and HCC.

Dataset Availability

The datasets used in this study can be obtained from the corresponding author upon reasonable request.

Getting Started

Prerequisites

• Python 3.10.4
• We recommend uv for fast environment and package management.

Installation

1. Create and activate a virtual environment:

   $ uv venv --python 3.10.4
   

   $ source .venv/bin/activate
   

2. Install the required packages:

   $ uv pip install -r requirements.txt
   

Training

The run.sh script is the main entry point for training and evaluation. You can customize the run by modifying the script or by passing arguments directly.

Parameters:

Flag	Description	Possible Values	Default (in run.sh)
-fold	Test fold for 5-fold cross-validation.	0, 1, 2, 3, 4	0
-bs	Batch size.	integer	8
-lr	Learning rate.	float	5e-5
-ep	Number of epochs.	integer	1
-wb	Log to Weights & Biases.	-	present
-wbStr	Name for the W&B run.	string	"evaluation run"
-net	Network architecture.	resnet50, xception	xception
-label	Comma-separated list of classes.	Leber_Parenchym, CCC, CRLM, HCC	Leber_Parenchym,CCC,CRLM,HCC

Example:

Execute the script to start a training run with the default parameters defined in run.sh:

$ ./run.sh

Repository Structure

• trainLoop.py: Main script for the training and evaluation loop.
• netModel.py: Defines the ResNet50 and Xception model architectures.
• CONFIG.py: Configuration file for paths, labels, and model parameters.
• argParser.py: Handles command-line argument parsing.
• LiverOCT2DZipDataset.py: PyTorch Dataset class for loading data from a ZIP file.
• run.sh: Shell script to execute training runs with different parameters.
• requirements.txt: A list of all python dependencies.

Citation

If you find this code useful for your research, please cite our paper:

[TODO ADD CITATION]