PS: my apologizes for the dramatic title and content, as already mentioned most of the segmentations are great and for the minority of others the cases are so extreme that i thought an easy fix would be provided by experienced users. At no point did i intend criticizing SPM which is a software of reference. I thought this was obvious but the first answer i received let me realize this could be misinterpreted.

On Wed, Oct 23, 2013 at 7:23 PM, Michael <[log in to unmask]> wrote:

Out of 42 T1 datasets processed with New Segment (SPM8, default options), 36 provided great segmentations and 6 resulted in Epic Fail.

Why Fail? Segmentations were not correct.

Why Epic? Tissues classes covered most regions OUTSIDE the brain/skull (and almost excluded brain/skull altogether)

Help? This is particularly "interesting" that the tissue maps are all in the background (i.e. in regions with intensity close to 0) so hopefully an easy fix exists. Any help would be greatly appreciated.

An example of T1 image for which the segmentation goes bad can be found there: https://www.dropbox.com/s/9zc5xvktxvqq8sp/T1.nii