AFNI version info (afni -ver):
macos_13_ARM_clang: Apr 4 2024 (Version AFNI_24.1.00 'Publius Septimius Geta')
hi all
just a general question without probably a perfect "right" answer but i'm curious about head motion at the group level (could apply to programs other than 3dLMEr). at the 1st level i 'cleaned' my fmri decently: motion params and their derivs regressed out, some censoring too, however i'd like to add at least one head motion covariate to regress out at the 2nd level as well. right now i'm using "average censored motion" but obviously there are a lot of other options i could pull from an AP output. what do you recommend or what have you seen other groups using? was thinking of also "max motion displacement" for example. i don't think it's wise to pick a number dependent on the scrubbing itself (like "num TRs above mot limit") and i assume it doesn't matter for regressing out if they're correlated? thanks!
Hi Sam,
Given that you are censoring, "average censored motion" makes good sense. But it also depends on the purpose of the covariate, including what the covariate is supposed to capture. And be careful, @Gang might have a lot to say on this!
While max displacement suggests motion, a high value does not necessarily mean that a subject was moving so much during the runs. Rather, a high displacement is what we would watch for when concerned about distortion changes over time.
Indeed, head motion is notoriously challenging to manage in fMRI data, with many factors at play. Its impact can vary significantly between task-based and resting-state data, and whether the motion is stimulus-induced makes a substantial difference. As you mentioned, there are some ad hoc approaches to addressing head motion, such as censoring and adding estimated head displacements as covariates at both the individual and group levels.
However, when we carefully consider the causal relationships involved, it becomes clear that while these methods may mitigate some of the effects, they also come with potential downsides. Conventional head motion adjustment methods can introduce biases that may distort or obscure the very relationships one is trying to uncover.
Ultimately, there’s no one-size-fits-all solution in fMRI data analysis, especially regarding head motion adjustments. While such adjustments (e.g., censoring, adding covariates) are often necessary as conventional steps, they can inadvertently undermine the original intent and lead to unintended distortions in the data.
thank you so much Rick and Gang for your super valuable food for thought. we'll be considering all this when finalizing what to include (and perhaps also how to respond to reviewers, fingers crossed)
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.