Brain MRI Segmentation

A hybrid framework combining GAN-based anomaly detection with U-Net segmentation for brain tumor detection in MRI scans. This project achieves high accuracy with limited training data by leveraging adversarial learning techniques.

Key Features

• GANet-Seg architecture for efficient tumor segmentation
• Achieves 88.84% Dice coefficient on BraTS dataset
• Works effectively with only 300 training volumes
• Implements advanced data augmentation techniques

Technologies

• PyTorch
• Medical image processing libraries
• GAN and U-Net architectures
• Python scientific computing stack