recon-all parcellation failure


I have been using -recon all for brain parcellation, it works well until some wrong parcellation in one of the healthy subject.
We made ROI maps by extracting the following anatomical areas based on the parcellation:
STS - ctx_lh_S_temporal_su
STG - ctx_lh_G_temp_sup-Lat
MTG - ctx_lh_G_temporal_mid
In this subject, the red and orange parts highlighted below are thought to be STG/S areas, but actually the yellow parts (I drew in ppt) should be the real STG/S.

I have tried to reconstruct a few times (and with different computers), the parcellation was still the same.
Here are the steps for recon-all (an automated script that I use for every subject):
cd …
set p = pwd
set SUBJECTS_DIR = {$p}
setenv SUBJECTS_DIR {$p}
recon-all -autorecon1 -s fs
recon-all -autorecon2 -s fs -notessellate -nosmooth1 -noinflate1 -noqsphere -nofix -nofinalsurfs
-nosmooth2 -noinflate2 -nocortribbon
recon-all -s fs -hemi lh -log $SUBJECTS_DIR/fs/scripts/recon-all_lh.log -status $SUBJECTS_DIR/fs/scripts/recon-all-status_lh.log
-tessellate -smooth1 -inflate1 -qsphere -fix -finalsurfs -smooth2 -inflate2 -cortribbon -autorecon3 &
recon-all -s fs -hemi rh -log $SUBJECTS_DIR/fs/scripts/recon-all_rh.log -status $SUBJECTS_DIR/fs/scripts/recon-all-status_rh.log
-tessellate -smooth1 -inflate1 -qsphere -fix -finalsurfs -smooth2 -inflate2 -cortribbon -autorecon3

What should I do in this case to get the correct STG/S ROI?



Hi, Yue-

I think that question should be asked to the FreeSurfer developers, who have written and maintain “recon-all”.