I am trying to utilise an analysis that includes the white matter, CSF and GS regressors from the fMRIoutput in a GLM, together with the denoised non-aggressive ICA-AROMA output. Given that the former regressors are calculated before the denoising from the ICA-AROMA occurs, I wanted to recalculate them, on the ICA-cleaned data.
Would that be simple in principle? Is there a way to do that through fMRIprep? I saw that this is normally calculated with a confounds.py script but I am not sure if this is applicable to my case and I was wondering if anyone in the community has developed a script for what I am trying to do. I plan to use FSL to do the first and second level analysis of my data.
To my understanding, you would instead want to regress the AROMA components and your other regressors of interest at the same time from a non-aroma output. That is, take your non-AROMA fmriprep output, and in your regression step include all confounds with aroma_motion_XX + your regressors of interest. See this note from Outputs of fMRIPrep — fmriprep version documentation
Maybe doing regression as normal on the AROMA output would be okay, but to me it seems cleaner to take a non-smoothed non-AROMA output, do all of confounds regression in a single step, and then do more things like smoothing. It seems like the regressor recalculation is a complicated work in progress, but using your confounds as is should suffice.
Thanks a lot for the reply. After looking a lot at other posts in neurostars (https://neurostars.org/t/fmriprep-ica-aroma-filtering-including-wm-csf-etc-confounds-in-fsl-regfilt/3137/2 I had ultimately decided to recompute the regressors from the cleaned data mainly because fsl crashes if I put too many regressors due to RAM issues (so adding that many components of the ICA might be an issue). Also that would constitute as aggresive AROMA right (there will be shared variance between the components and the regressors that I want to add (CSF, WM)?
Just to clarify I would simply want to recalculate the regressors for CSF, WM and GS from the cleaned-ICA ARoma output (in order to only add those in the GLM)
