Hi Freesurfer team,

I am trying to extract ROI PET data from the cortical regions using the Desikan atlas, and I am wondering which is the recommended approach.

1) mri_segstats on resampled aparc+aseg.mgz in PET space or 2) mri_segstats on lh.PetSurface.mgh in native surface (not fsaverage).

Option (1) always lead to to substantially lower BPND values in some regions compared to the measures derived on the surface.  If I would like to take the cortical measurements from lh.PetSurface.mgh, should I smooth the data first?

A second question relates to projfrac.  I am working with tau BPND data and there is quite a bit of CSF smearing in the outer layers of the cortex, resulting in negative BPND voxels. However, even without CSF contamination, there is quite a steep drop off in BPND intensity from the WM boundary to the pial surface (projfrac 0.5). The registration is OK too. For my data, sampling nearer to the WM surface (0.1) seems to result in positive BPND voxels compared to projfrac 0.5. As such, is it OK to use projfrac 0.1 for all subjects in my study?

Thanks for taking the time to read this!

Best Wishes, Elijah