We've using version 5.1 to conduct a longitudinal study involving over 150 subjects with up to three time points apiece, and we're having some issues with regards to aseg in the areas around the cerebellum. Attached is an image of the problem - some dura or something else is being classified as cerebellum. It's particularly troubling because this issue is present in most of our images, but to varying degrees. We've tried adjusting the watershed parameters (using multistrip) as well as using gcut, but neither method was effective in removing the offending voxels.

 

It seems like our only other option is to manually edit the brainmask.mgz, but with so many subjects and time points, we're wondering if there is another automated fix that we haven't tried. Otherwise, is this something that we'll have to manually fix, or can it be left alone?

 

Thank you,

 

Michael Murray