Hi Lijie (cc to mailing list),
I wrote a script to do just this in July. Here is a link to a tar file. It requires knowledge of matlab and bash, but is pretty straight forward. It
https://www.dropbox.com/s/s1f78j2ozv1gw8q/subfield_to_ROI.tar.gz
a) transforms the subfield probabilities to nu.mgz space (bash, using mri_label2vol) b) applies a threshold to the maps (Koen suggested 50% of max probability, hence 128 of 255) to avoid over-estimation of the subfield volumes (bash) c) comparing the probabilities of individual subfields in IJK space and assigning a unique label to each subfield
We can continue this discussion off-line if you like
freesurfer@nmr.mgh.harvard.edu