Dear FreeSurfers,
I'm working on a structural dataset and after running recon-all -all, I had to correct all cases (32 so far) from minimum pial/wm corrections to massive pial corrections (only three cases needed control points). Mostly, the corrections were needed due to remaining dura/skull. Initially, I tried to adjust the skull-stripping (either with changing the wsthresh or gcut for individual cases), however, those corrections were not always too successful (i.e. still remaining dura/skull), so I continued correcting manually.
Are there any known/expected problems that could still arise with the aparc+aseg segmentations after (massive) corrections (i.e. all cases), even after redoing respective recon-steps with version 5.1? In the release notes, I only found that control points may adversely affect aseg... I checked the segmentations and so far I couldn't recognize any obvious/consistent flaws. I remember from working with version 4.1. that most cases need at least some corrections, but I don't have any experience with 5.1, yet.
I'm specifically interested in the amygdala, but of course amygdala segmentations will be corrected manually.
Thanks, Dorit
Data were acquired on a 3T Tim Trio dataset (32 channel-coil, TR 1900ms, TE 2.52 ms, 1x1x1 voxel-size, TI 900ms, flip angle 9, ipad grappa2, bandwidth 170Hz/pixel, prescan-normalize). Data were initially processed with recon-all -all with version 5.1 on a linux cluster (scientific linux 6.1). Data were corrected and then re-run with respective recon-steps on a local Mac (OS 10.5.8) with freesurfer-Darwin-leopard-i686-stable-pub-v5.1.0.