Besides trimming images, to exclude the neck and empty space, what else have people tried to improve segmentation quality? As others have noted, some images are improperly segmented, such that gray matter partitions contain all tissue types, while WM and CSF are underestimated. These images do not substantially differ from those that do segment properly (e.g. atrophy, contrast). I have observed this problem both for intact as well as skull stripped images (using a stripped T1 template). Many thanks, Audrey Audrey Duarte, Ph.D. Postdoctoral Researcher San Francisco VAMC 4150 Clement St, Mail stop:116R San Francisco, Ca 94121 415-221-4810 x 3852