We have assessed freesurfer output for 160+ brains, and about 50% have poor segmentation of pial matter that includes too much skull.  Manual edits and re-running does work to improve the problem, but is very time/labor intensive for so many scans.

I saw on the freesurfer wiki that you can include T2 or Flair images to improve the segmentation for this problem. However, we do not have these extra scans. What do you recommend to do otherwise? 

Thank you!

--