Multiple sclerosis detection from MRI images.
- Dataset from eHealth lab of Department of Computer Science at University of Cyprus.
- Pre-processing: apply continuous or discrete wavelet transformations to MRI images.
- Binary classification with CNN: predict whether the input image is healthy or unhealthy using CNN with AlexNet or SqueezeNet macro-architectures.
- Wavelet Transformations: (ipynb)
- Batch Wavelet Transformations: (ipynb)
- AlexNet results: (AlexNet ipynb)
- SqueezeNet results: (SqueezeNet ipynb)
val_loss: 0.4376 - val_accuracy: 0.8297 - val_mse: 0.1396 - val_mae: 0.2852