Hello fellow Freesurfer user,

I am currently working on an ageing data set (~60 subjects between 19 to 81 years old). I mainly want to use Freesurfer for two different things:

1. get a cortical and sub-cortical segmentation to use those regions for a connectivity analysis
2. use the WM/GM boundary and the pial surface to compute cortical thickness
   

I have run recon-all -autorecon2 for all subjects based on T1-weighted structural images. I am no in the process of making sure that everything worked before continuing (I already checked after autorecon-1 that the skull striping didn't remove large chunks of the brain). Therefore, I looked at the surfaces and saw that there are problems with almost half of my subjects due to bad skull striping (WM/GM boundary or pial surface extends to skull). I also read that lesions (WM surface cutting into the brain instead of following it) can lead to dimples or holes in the inflated surfaces. I know that one can correct some of those errors manually, but probably not all the errors will have an impact on what I want to use it for. Which brings me to the questions:

1. Which parts are critical for my subsequent analysis (concerning points 1 and 2 above) and therefore need to be corrected?
2. How much time should you generally invest on manual intervention (erasing voxels in brainmask.mgz, filling holes in wm.mgz, etc.) and when should you rather refine parameters in the recon-all processes (see also 3)?
3. If there are a lot of errors due to wrong skull stripping, how can you change steps in the pipeline to improve it?

I would really appreciate help with any of the three points, so feel free to just answer to a subset of them. Many thanks in advance!

Best regards,
Niklas