I have followed your advice to see whether I can segment the brainstem into midbrain, pons and medulla
successfully. The template in standard space I used is called wfu_pickatlas, which has segmented the brainstem
into midbrain, pons and medulla. Then I registered the T1 image of a TBI patient to standard space. I checked
the registeration. It looked fine. Then I inverted the warp and transformed the atlas labels to the T1 image.
As you suggested, what I need to do next is to mask them with the brainstem segmentation from FIRST.
The problem I got is that the two brainstem structures (one from FIRST and another from registeration as above)
didn't fully overlap. The differences are quite big. I think I can't use the borders to segment the brainstem.
All the best