Thank you very much for the help. We could solve the problem with that subject but now we have a similar problem with another subject. In the previous case, we have only one problematic voxel in the skull, but now the label 75 (it shouldn't exist either) is taking a lot of voxels in the ventricles and the caudate. So, I think that we should use other solution to fix it.

I know that this issue is weird. We use gtmseg command with more or less 500 subjects and we only have this type of problem with two subjects.

Best regards,

Marina.

 > Probably the easiest thing is just to edit that voxel  to be one of the neighboring 

 > segments. First, load the aseg and find that voxel. See which segment(s) are around 

 > that voxel and get the segmentation id. Then run
 > mri_binarize --replaceonly 229 YourNewSegID --i aparc+aseg.mgz --o 

 > aparc+aseg.mgz

 >  then re-run gtmseg