Brain Tumor Classification Using Graph CNN and Deep Learning

Overview

Brain tumors are among the most critical neurological disorders, requiring accurate and timely diagnosis for effective treatment planning. Magnetic Resonance Imaging (MRI) plays a vital role in detecting and classifying brain tumors.

This project presents a Graph-Enhanced Convolutional Neural Network (Graph CNN) framework for automated brain tumor classification using MRI images. The proposed approach combines image preprocessing, data augmentation, graph-based feature representation, and deep learning techniques to improve classification performance and support AI-assisted medical diagnosis.

---

Key Features

• Multi-class Brain Tumor Classification
• MRI Image Analysis
• Graph-Based Feature Representation
• Convolutional Neural Network (CNN)
• Data Augmentation
• Medical Image Processing
• Healthcare AI Application
• Automated Diagnostic Support

---

Dataset

Brain Tumor MRI Dataset

The dataset contains MRI brain scans categorized into multiple diagnostic classes.

Classification Categories

• Glioma Tumor
• Meningioma Tumor
• Pituitary Tumor
• No Tumor

Dataset Characteristics

• Brain MRI Images
• Multi-Class Classification
• Medical Imaging Dataset
• Suitable for Deep Learning and Healthcare AI Research

---

Methodology

Image Preprocessing

The following preprocessing techniques were applied:

• Image Resizing
• Normalization
• Noise Reduction
• Contrast Enhancement
• Data Cleaning

Data Augmentation

To improve model robustness and reduce overfitting:

• Rotation
• Horizontal Flipping
• Zooming
• Shifting
• Scaling

---

Graph-Based Learning

Unlike traditional CNN approaches, this project incorporates graph-based feature extraction techniques.

Graph Construction

The MRI image features are transformed into graph representations where:

• Nodes represent extracted image features
• Edges capture feature relationships
• Adjacency matrices model structural dependencies

Benefits

• Captures spatial feature relationships
• Improves feature representation
• Enhances classification capability
• Supports graph convolution operations

---

Deep Learning Architecture

Graph CNN Framework

MRI Images
      │
      ▼
Image Preprocessing
      │
      ▼
Data Augmentation
      │
      ▼
Feature Extraction
      │
      ▼
Graph Construction
      │
      ▼
Graph Convolution Operations
      │
      ▼
CNN Classification Layer
      │
      ▼
Tumor Classification

---

Technologies Used

Programming Language

• Python

Deep Learning

• TensorFlow
• Keras

Computer Vision

• OpenCV
• Pillow

Data Science

• NumPy
• Pandas

Visualization

• Matplotlib
• Seaborn

---

Evaluation Metrics

The model performance was evaluated using:

• Accuracy
• Precision
• Recall
• F1-Score
• Confusion Matrix
• Classification Report
• ROC Analysis
• AUC Score

---

Experimental Workflow

1. Dataset Collection
2. Image Preprocessing
3. Data Augmentation
4. Feature Extraction
5. Graph Construction
6. Graph Convolution Operations
7. CNN Training
8. Model Evaluation
9. Performance Analysis

---

Results

The proposed Graph CNN framework demonstrates the effectiveness of combining:

• Medical Image Processing
• Graph-Based Learning
• Deep Learning
• Computer Vision

for automated brain tumor classification.

The framework successfully distinguishes between different tumor categories and healthy brain scans, supporting intelligent healthcare applications.

---

Results Visualization

After uploading your result images, include:

Confusion Matrix

![Confusion Matrix](images/confusion_matrix.png)

Training Performance

![Training Curves](images/training_curves.png)

Classification Results

![Prediction Results](images/predictions.png)

---

Repository Structure

brain-tumor-classification-using-graph-cnn/

├── FinalBrainTumorClassification.ipynb
├── README.md
├── requirements.txt
│
├── images/
│   ├── confusion_matrix.png
│   ├── training_curves.png
│   ├── predictions.png
│   └── sample_mri_images.png
│
└── dataset/

---

Requirements

The project was developed using:

• Python 3.x
• TensorFlow
• Keras
• OpenCV
• NumPy
• Pandas
• Matplotlib
• Seaborn

Development Environment

• Jupyter Notebook
• Google Colab

Hardware

• GPU acceleration recommended for faster training
• Compatible with Google Colab GPU environments

---

Applications

Healthcare AI

• Brain Tumor Diagnosis
• Medical Image Classification
• Clinical Decision Support Systems
• Computer-Aided Diagnosis (CAD)

Research

• Medical Imaging Research
• Deep Learning Research
• Graph-Based Learning
• Healthcare Analytics

---

Future Work

• Graph Neural Networks (GNN)
• Vision Transformers (ViT)
• Explainable AI (XAI)
• Attention Mechanisms
• Real-Time Clinical Deployment
• Multi-Modal Medical Diagnosis

---

Author

Ankur Ray Chayan

Machine Learning Researcher | Embedded Systems Researcher

Research Interests

• Artificial Intelligence
• Deep Learning
• Computer Vision
• Medical Image Analysis
• Healthcare AI
• Graph-Based Learning

GitHub: https://github.com/AnkurRay25

ResearchGate: https://www.researchgate.net/publication/385551892_st-2018-d0100-l_10-1055_s-0037-1609490-1

---

License

This project is licensed under the MIT License.

---