We're working using vbm8 (that uses DARTEL for normalizing) in order to segment human 3D T1-weighted images.
Looking at the results we notice that the subcortical structures (thalamus, caudate nucleus) are oddly hyperintense, i.e. the probability of these structures to belong GM is much higher than the cortical ones.
Is there a known problem with the segmentation process?
Thank you in advance
Greetings
|