> The manual page for FNIRT says that it is diffeomorphic, i.e. in a mapping from U to V, every voxel in U has a corresponding position in V.
Like with most "diffeomorphic" methods there is still a small risk of a non-diffeomorphic mapping. But yes, for all practical purposes it is diffeomorphic.
> So after using FNIRT can I know where a given voxel (x_u,y_u,z_u) in U ended up in V? Corresponding voxel coordinate/index in V?
If you have warped U to V you need to use invwarp to invert the warp field in order to find where every voxel in U ended up in V.
Good luck Jesper