Let me start by saying thanks for your making your code available.
My question is: can you please provide the raw feature matrix you use for training and also the code used for feature extraction? My understanding is that you use the texton filters on each volxel so I imagine that the feature matrix would be huge given the number of examples you must have used. Providing a download link would be great.
Thank in advance.

Kind Regards

If you go to the downloads page and download W2MHS v1.1 Source and Training Data it will contain our feature matrix. It is about 1GB. I wouldn't use the code that comes with it but you can find the features_training.mat file in "/W2MHS/training/"

Chris