Consider 'hiding' the spinal cord cropping inside the MS lesion segmentation task

[jcohenadad]

I am afraid that some users won't crop the spinal cord and will try to apply the MS lesion segmentation model directly on the full image, which would produce bad results.

Should we consider 'enforcing' spinal cord cropping, ie: doing a cascaded model of: SC segmentation & MS lesion segmentation?

We could still keep the SC cord segmentation model as as separate model, in case users want to only segment the cord.