Following Moskovitch et al..
I guess that right now our approach would be: "Calculate these things separately and point pyAFQ to the derivatives", so arguably out of scope for pyAFQ itself.
Maybe it needs to be a separate project? Or is it part of qsirecon?
Tagging @mattcieslak who also wants to make this happen.