I am working on a data set and am having particular difficulty with one subject. It appeared the skull strip was mistakenly removing occipital regions. These regions were absent from the brainmask, T1, and resulting surfaces. I attempted to solve this in 2 different ways. I first adjusted the watershed and reran the recon-all -autorecon2 in order to redraw the surfaces. Adjusting the watershed worked to include these regions in both the brainmask and T1 but the resulting surfaces hugely inaccurate (see the attached screenshot  '4003_badsurfs_parietal_cerebellum2'; Oddly enough, the frontal surfaces are largely accurate but the parietal regions include gray matter in the white surface and even large amounts of the cerebellum, while excluding occipital regions.

I then reran it from the raw dicoms and had the same issue with occipital regions being excluded so I adjusted the watershed and tried to manually fix the surfaces, but this resulted in similar inaccurate surfaces. See the second attached screenshot '4003_badsurfs'. ll surfaces are fairly messy, but some regions are more accurage, where the appears to distinguish between gray and white and other areas are over inclusive, but cerebellar and some superior and medial regions are excluded. 

Any and all help would be much appreciated. Thanks in advance!

Kate