Thanks for your response. Actually, we do modify the aseg.mgz (Sorry, I didn't realized and I shouldn't have considered this a 'standard FS pipeline').

We work with Multiple Sclerosis patients and because of the size and number of lesions in some of them, the pial and WM surfaces that we obtain after a first run of “recon-all -all ” are not accurate. FreeSurfer pipeline for healthy subjects or small lesions requires to edit wm.mgz -- but these edits don't fix our issue of WM & GM surface getting pulled in by lesions. Given the size and number of lesions in some of our MS patients, using control points make the entire recon worse. So after repeated testing, the solution we arrived at was to insert manually drawn lesion masks into segmentation volumes as "WM-hypointensities". We label manually drawn lesion masks as "WM-hypointensities" on aseg.mgz, and wm.mgz. We then do a second run of recon-all and check the surfaces again. Once they are ok we run “recon-all -autorecon3” and check the final output.

In principle, we only change voxels in aseg.mgz to "WM-hypointensities", so I don't know why these unexpected values appear. Just to check, I renamed the 'aseg.auto.mgz' as 'aseg.mgz' and re-run again the steps for PVC:

1. Segmentation for GTM:

gtmseg --s $subj"_recon" --xcerseg

In this step I get the following warnings:

2. Registration:

mri_coreg --s $subj"_recon" --mov $path_to_suvr/SUVR.nii.gz --reg $path_to_suvr/SUVR.reg.lta

3. PVC:

mri_gtmpvc --i $path_to_suvr/SUVR.nii.gz --reg $path_to_suvr/SUVR.reg.lta --psf 6 --seg gtmseg.mgz --default-seg-merge --auto-mask PSF .01 --mgx .01 --o $path_to_recon/$subj"_recon"/gtmpvc.output.

And still get the same error (below) (maybe because there are other files that are based in the previous aseg.mgz? )

Any help with this would be very much appreciate it!

Thanks!

Elena