Hi-
The fat_proc_map_dti program uses affine registration, so bringing data from a subject's own anatomical dataset to the DWI that has been aligned to it makes sense with that level of warping.
To bring ROIs from a reference template to a subject DWI/DTI dataset would likely require nonlinear alignment for reasonable accuracy; once the transformation/warp has been estimated, one can apply the warp dataset to the ROIs, bringing them into the subject anatomical space.
Reference templates are typically T1w volumes. Do you have the subject's T1w anatomical dataset in DWI space, like from using fat_proc_map_dti? If so, you could perform nonlinear alignment to that with the default cost function (lpa+ZZ), e.g., using @SSwarper or 3dQwarp. If you want to align a template to a subject's b=0 DWI volume or to a T2w volume, you could use 3dQwarp or @SSwarper, but use a cost function that will be OK with different dataset tissue contrasts---say, lpc+ZZ.
--pt
