How to deal with behavioral variables correlated with mean framewise displacement (FD)?

Dear all,

As the requirment for resting-state fMRI studies on controlling head motion is becoming stricter and stricter. Typically we have to use a more stringent criteria to exclude participants for the head motion. For example, conducting scrubbing in the preprocessing and excluding participants who have large mean framewise displacement (FD) (e.g., FD > 0.2, 0.15). And we also need to make sure that our behavioral variables are not correlated with the mean FD, to convice the reviwers that our results are not influenced by the head motion.

However, now I am facing a problem where my behavioral variables are correlated with the mean FD after I excluded participants with mean FD > 0.2mm (Jenkinson). One paper (Rosenburg, 2015) tried to use a more scricter criteria, for example, excluded participants with mean FD > 0.06 mm and then finally their behavior was not correlated with the mean FD. In my case, I tried to use mean FD > 0.15 and it almost excluded half of the total sample. However, even after this, the mean FD is still correlated with the behavioral. If I adopted a more stringent criteria, say > 0.1, then I am going to exclude 70% of the total sample, which doesn't make sense to me to continue the analyses with only 30% of my data.

I am wondering how to deal with this issue? I think it might be helpful if I regress out the mean FD in my image data, but not sure if it is a reasonable way to do so.

Any suggestions would be very appreciated! Thank you so much for your attention!

Best wishes,

Mengxia