Should I calculate the CSF, WM or GS nuisances again when I regress them out from fmriprep results?

Hi all,
After fmriprep, I want to process my results furthermore including smoothing, regressing out nuisance(mean white matter and CSF), and filtering.

I try to use “3dTproject” to regress out those nuisances and filter simultaneously.

I found mean white matter and CSF signals in fmriprep output results(xx_desc-confounds_regressors.tsv), I think they are extracted from BOLD images based on ROI (xx_label-WM_probseg.nii.gz and xx_label-CSF_probseg.nii.gz) of individual native space. Is it right?
Could I regress out them on results of standard space directly? Or Is it better for us to extract them in standard space based on a standard template such as SPM’s CSF template or MNI152NLin2009cAsym WM mask? These WM, CSF, and brain masks of different standard spaces could be obtained in OSF | TemplateFlow. But I think it’s difficult to find those masks of standard space if we process surface files (i.e., *.gii, *.dtseries.nii). Any suggestions?
Thanks a lot!

1 Like

I don’t think there’s any need to recalculate your regressors. Also, if you’re interested in denoising/filtering fMRIPrep derivatives, may I suggest using XCP_ABCD https://github.com/PennLINC/xcp_abcd? It has widely used denoising pipelines built in to it.

Steven

1 Like