3D Brain Tumour Segmentation

It is federated framework for the segmentation task, which uses novel loss and aggregation strategies.

This project implements a 3D U-Net model for medical image segmentation using TensorFlow and the segmentation_models_3D library. The project is structured into multiple modules for better maintainability and readability.

Project Structure

data_loader.py: Contains the DataLoader class for loading and preprocessing the dataset.
model.py: Contains the UNET3D class for defining the 3D U-Net model.
train.py: Contains the Trainer class for training and evaluating the model.
utils.py: Contains utility functions for metrics and loss calculations.
main.py: The main script to run the training process.

Requirements

Python 3.6+
TensorFlow 2.x
segmentation_models_3D
numpy
tqdm

You can install the required packages using the following command:

pip install -r requirements.txt

Usage

Clone the repository:

https://github.com/SUNNY11286/SAFCF-federated-contrastive-segmentation-framework.git

Run the training script:
```
python app.py
```

Customization

DataLoader: Modify data_loader.py to change how data is loaded and preprocessed.
Model: Modify model.py to change the architecture of the 3D U-Net model.
Training: Modify train.py to change the training process, including the loss function, metrics, and training loop.

Contributing