I am doing a surface based ROI analysis where for each subject, isomorphic ROIs of 300 vertices (centered around peak functional activations) were found on the fsaverage surface space. I was wondering if I should register these masks that I have created back to each subject's individual surface before creating labels for these ROIs, and if so, what would the registration file be to register the fsaverage space back to each subject's space? The goal of this is to be able to run mri_glmfit on just the spaces within these labels. Is there a better method? Also, each subject will have a different number of labels considering that not every ROI was present on every subject. will this be a problem when running glmfit?