Add lesion mask argument to antsBrainExtraction

[dorianps]

Currently antsBrainExtraction.sh uses SyN registration This creates some problems in brains with stroke (or other lesions), because the deformation of the lesion causes some of the brain mask to go in the lesion. Sometimes is more evident but here is a more subtle example:

I am thinking that is might be a good idea to add an argument for a lesion mask, and then remove the mask from registrations. I am not that practical at implementing this without major changes in the script (i.e., I usually use the mask on the fixed image, but here the lesioned image might be the moving image), but I am sure that such option would be helpful to all those who work on lesioned brains.

[ntustison]

Can you post data for me to play with? Assuming that the user would have to supply the initial lesion masks for a given cohort, I don't know if that would be less time-consuming than simply correcting any errors in the brain extraction masks post-processing. Perhaps a better route would be to allow for increased regularization in the registration, e.g., BSplineSyN.