We have a T1 image that has been problematic to process. This patient has thicker skull and extensive hypointensities in the white matter along with big vents.
Skullstripping by modifying the watershed parameter either leaves too much skull or removes entire chunks of the brain in this Siemens image; unfortunately there seems to be no middle ground.
In recon-all FreeSurfer has been unable to discern the gray matter boundary in the posterior of the brain and omits folding detail of the sulci. I am attaching a screenshot overview of the "best" attempt so far, which finished this morning (there is a bit of skull left over on this run; on previous runs where the skull is cleanly removed this problem persists). As a last-ditch effort I had re-sliced the axis-reformat T1 images as additional images (002.mgz, 003.mgz, 004.mgz) (any feedback on the feasibility of this approach is welcomed also).
Any suggestions to improve the processing on these cortical problem areas are very much appreciated (ventricles are as good as I can hope for today; that's matter for a whole other post!).
Much thanks,
Eve LoCastro Senior Research Assistant IDEAL, Radiology Department Weill Cornell Medical College 515 E. 71st St S118 Tel: 212-746-1289 Fax: 212-746-4189 http://ideal-cornell.com/http://ideal-cornell.com
