I may still not be clear on what you're attempting to do. You can certainly segment the brain in GM/WM/CSF in AFNI. But if you want to do inverse calculations then you probably want to stick to an EEG/MEG software like MNE/FieldTrip/Brainstorm. I'm most familiar with MNE, which can fit your BEM source models to individual subjects and then export those to GIFTI for overlay in AFNI/SUMA. You could then use a combination of
3dSurf2Vol,
MapIcosahedron, and
SurfToSurf to map data from subject space to group space.
-Peter