Hello,
We have some signal dropout at the top of the brain in some scans, possibly similar to Rutvik Desai's 02/18/10 post. However, in our case, FreeSurfer is not including any of the regions affected by the dropout in both its subcortical segmentation and cortical parcellation. Instead, FreeSurfer essentially begins to treat the gray/white boundary as the CSF/gray surface, and classifies white matter as gray matter. The severity of this issue seems to correlate with how severe the dropout is.
I have tried running nu_correct (normalization) for up to 256 iterations. This produces a noticeable effect when visualizing the volumes in tkmedit, but has no effect on FreeSurfer's output. I can upload a sample subject, if desired. We are using FreeSurfer 5.0 x64 on a Sun Grid Engine operating servers running CentOS 4.5.
Thank you, Victor Yee