annemsony / gnn

working on Graph Neural Network

GNN for Alzheimer's disease prediction from Diffusion Tensor Imaging Data

This project involves the implementation of a Graph Neural Network (GNN) to predict labels from a dataset derived from diffusion tensor imaging (DTI) data. Specifically, we aim to classify the graph structure into one of four categories: AD, CD, Early Detection, and Late Detection.

Project Overview

The dataset used in this project consists of adjacency matrices, each representing a patient's DTI data. Our goal is to leverage these matrices to predict the corresponding label using a GNN.

Key Features

• Task: 4-class classification (AD, CD, Early Detection, Late Detection)
• Input Data: Adjacency matrices derived from DTI data
• Output: Predicted labels for each graph

Data Preparation

Each patient file is represented as an adjacency matrix. To process this dataset with a GNN, the adjacency matrix needs to be converted into a graph structure, comprising:

• Nodes
• Edges
• Weights

These components are extracted from the adjacency matrix to create the input for the GNN.

Implementation Details

• Model: A GNN model tailored to handle the graph structure of the dataset.
• Classification: The model is trained to classify the graph data into one of the four classes: AD, CD, Early Detection, or Late Detection.

Getting Started

Prerequisites

Ensure you have the following installed:

• Python 3.x
• PyTorch
• DGL (Deep Graph Library) or any other GNN framework of your choice
• Additional libraries as required (NumPy, SciPy, etc.)

Installation

Clone the repository and install the required dependencies:

git clone https://github.com/your-username/gnn-dti-classification.git
cd gnn-dti-classification
pip install -r requirements.txt