Head motion estimation on DWI is poor

[oesteban]

We're using 3dVolreg (from AFNI), copying the procedures from BOLD counterparts. In the long term, I think education should do this better than anything else, but in the short term, we may want to talk with Paul Taylor and find a good configuration of AFNI for DWI.

[arokem]

In the long term, I think education should do this better than anything else

Do you mean "eddymotion"? I'd think that would still be pretty time-consuming to run in most case, beyond what you'd want here?

Might also want to try dipy.align.motion_correction

[oesteban]

Do you mean "eddymotion"?

hehehe yes :D

I'd think that would still be pretty time-consuming to run in most case, beyond what you'd want here?

I think we want to get it in good shape and have simple models that are competitively fast.

Might also want to try dipy.align.motion_correction

Sure! I'd be happy to try it.