External Email - Use Caution
Hi everyone,
I am processing the NIFD dataset (visit 1) and the FS (v 6.0) segmentation consistently includes dura mater in the pial boundary in most subjects (I could not attach example snapshots to this email, but they are available if required).
I have tried -gcut and the watershed algorithms for FS and they do not seem to be making a difference. Considering that the default preflooding height in FS is 25, I chose 20 and 1 for additional processing. While PFH=1 (and -gcut) does not remove the skull left behind (just shaves off some boundary pixels), using PFH=20 (oddly) seems to actually add more pixels. These observations are based on the brainmask.mgz output after running “ recon-all -skullstrip -wsthresh <pf height> -clean-bm -subjid <subid>” and “recon-all -skullstrip -clean-bm -gcut -subjid <subjid>” (and without fully reprocessing the data).
- I would be very grateful for any insights about additional methods to try to remediate this problem, short of manual editing. This is a widely occurring issue in the dataset, so we cannot just exclude the problem cases.
- Would it be advisable to use different software for the skull stripping (such as BET) and use that output in the FS pipeline?
- Could I be missing something (this is the first time I am using FS)?
Thank you very much for your time!
Best regards,
Divya