Refactor prepare_data.py

[DrAnaximandre]

should we break out and add comments to prepare_data.py ?

[DrAnaximandre]

Also i spotted l19 l20:

X_f = f.create_dataset("image", [total_slices,217,181,1], dtype='float16')

Y_f = f.create_dataset("label", [total_slices,217,181,1], dtype='float16')

I'm not sure it works fine when dealing with a feature dimension other than 2 ?

[DrAnaximandre]

Also lines 44 to 48 are to be removed (the index_bad_Y)