When doing multi-echo data preprocessing in AFNI (afni_proc.py example 12.c) with tedana, why do the examples use ‘-regress_apply_mot_types’ in the script?
My idea is that tedana would have removed all motion-related ICA components from the data - why is motion being regressed out again in AFNI?
Given our current understanding, I’d say using and not using are both fine choices. ICA components with substantial motion artifacts are likely to be removed within tedana, but tedana is regressing out time series of those ICA components. It is plausible that the 12 motion parameters and their derivatives will regress out additional signal, but at the cost of removing 12 more degrees of freedom.
We don’t yet have a good handle on what -regress_apply_mot_types removes vs tedana. There is definitely a lot of overlap, but I it’s probably not 100% overlap. This is something I and others are looking into. The tedana group is working on a new feature what will make it possible to directly fit motion parameters to ICA components. This will potentially make it possible to more fully capture the same information within tedana and to better understand the relationship between regressing out motion via ICA components vs using motion regressors directly on the data ( Generate metrics from external regressors using F stats by handwerkerd · Pull Request #1064 · ME-ICA/tedana · GitHub )
MANAGED BY INCF