Fmriprep: strange T1w segmentation with surface reconstruction

1

Dear all,

Is it expected to have a different T1w segmentation with surface reconstruction compared to without recon-all? I did not find an answer to this question despite a thorough search on neurostar.

For example, for the case below (elderly patient), I have a strange T1w segmentation with surface reconstruction with a less neat GM mask than without surface reconstruction. BTW, the surface reconstruction seems to me perfectly fine.
Is the strange segmentation more precise than the other, or is it better to run both segmentations with and without surface reconstruction for each subject and keep the latter segmentation?

I ran:
fmriprep-docker C:\Users\Utilisateur\Desktop\MRI_example C:\Users\Utilisateur\Desktop\MRI_example/derivatives --participant-label 01001 --output-spaces MNI152NLin6Asym:res-2 --fs-license-file C:\Users\Utilisateur\Desktop\MRI_example/derivatives\license.txt

dseg_recon-all
dseg_recon-all1920×821 515 KB

fmriprep-docker C:\Users\Utilisateur\Desktop\MRI_example C:\Users\Utilisateur\Desktop\MRI_example/derivatives --participant-label 01001 --output-spaces MNI152NLin6Asym:res-2 --anat-only --fs-no-reconall --fs-license-file

Thanks in advance for your help!