Hello, I hope this email finds you well.

 

I am using Samseg for the MS patients' data and am very impressed with its performance in terms of lesions segmentation along with posteriors and probability maps saving. Also, compared to freesurfer, it has a remarkably better quality of cortical gray matter segmentation, particularly in subjects with juxta/intracortical lesions. However, to my knowledge, it does not allow doing a cortex parcellation. I used it with the —recon option, but in that case, the output (aparc+aseg.mgz) is identical or very close to those freesurfer provides and commonly has some segmentation errors, whereas seg.mgz saved in samseg directory demonstrates significantly better segmentation.

I wonder if there is any way to run cortex parcellation using exact samseg seg.mgz output?

I have attached a couple of images demonstrating the difference of segmentation output for freesurfer and samseg.

 

Thank you,

 

Best regards

Valentin