Aligning to a previous run: Effects on motion regressors?


I just wonder if it matters what volume you align to when it comes to motion regressors. Assume we don’t do any warping, just the align block and for some reason (in this time real time fMRI) want to align the EPI volumes to an EPI volume from a previous run (localizer or seed mask from a pervious run). When preforming the regression of the current run with the motion regressors, will this mattter?

I base this on the assumtion that the motion regressors are generated during the alignment: The ammount you have to re-orient one EPI to the next gives motion and angulation in the 3 directions of space. If you then align to an EPI not from this run, will that cause problems?

I think it won’t since i guess that would also cause trouble with the “align to min outlier volume” but I just want to make sure!


Hi Robin,

It should make a difference, but we do not expect it to be big.
And actually, why would aligning to any particular volume be
necessarily better than another from that perspective?

What we hope is that alignment to a different volume mostly
amounts to a constant shift among all motion parameters,
akin to offsetting them by the parameter difference between
the base volumes in question. And a constant shift in them,
per run or across, would have no effect on them in the
regression (it would be absorbed by the constant polort

So again, there will be some difference, but it should be
small. The advantage of using the min outlier is that it
should be more robust, since that volume should at least
not be corrupted by motion.

  • rick

Thanks Rick!

But this is true even if the volume is from another run?

Then I can keep my current set-up which is first preforming a localizer task where the EPIs are volreged to the min outlier and then the anatomy is aligned to that epi volume. Then you draw a ROI of your choice on the anatomical . Then during the real time feedback run the EPIs are aligned to the minimum outlier epi of the localizer scan so that the ROI matches the same anatomical structure as you drew the ROI on. Motion effects on the real time BOLD are taken care of with 3dDeconvolve -x1D stop and 3dTproject using the motion regressors. Thanks, now I don’t have to worry!

Yes, it is expected to hold true regardless of the run.
What you are doing sounds quite reasonable to me.

  • rick