This code is used for project related to predicting Alzheimers using Capsule Network
We preprocessed brain MRI scans with segmentation, registration, normalization, noise addition and skull-stripping. Processing is done using MRI scan with segmentation, normalisation, augmentation and noise addition along with stripping.
- Trained using 3D autoencoder first on few patches extracted from scans to generate feature whiich are inputed to 3D CNN for predition
- This preprocessed image is fed as input in Capsule network for better prediction task.
Validation accuracy with pre-trained model is 91.3%