Hello,
Sorry for being so slow on this.
You mention is one volume of AP direction map and one volume of PA direction map for each run of the epi images. Those are opposite phase directions, they are not B0 distortion maps.
What we do in this case is to non-linearly register the forward direction (PA?) with the reverse direction (AP?) map, and then take the "midpoint" or the sqrt(xform). That can be done easily in afni_proc.py, but providing one pair of those images using -blip_forward_dset and -blip_reverse_dset. Very easy.
The difficulty here would be doing this per run. Right now, afni_proc.py is only able to do this once across all runs. You could work around this limitation by running AP per run, maybe giving it the first run's volreg base for the rest to align to, or you could just use the first pair only.
I will ponder this going forward.
- rick