Masks used for generating white_matter, csf, and global_signal confounds


I was wondering if anyone could answer a couple of questions I have about the generation of the white_matter, csf, and global_signal confounds in fMRIPREP.

(1) Are the native T1 masks found under /sub*/anat/*probseg.nii.gz what is used to define the different ROIs (with global_signal coming from union of WM, CSF, and GS)?

(2) These masks don’t appear to have any erosion. Is any mask erosion done prior to calculation of these signals? If so, how much is done and is the eroded mask stored anywhere?

(3) These signals haven’t been temporally filtered at all, right?

Thanks in advance!!

Hi @erik_lee,

I would like to follow up on your question. I am also wondering if the WM and CSF masks provided by fmriprep (sub-<subject_label>[_space-<space_label>]_label-WM_probseg.nii.gz) are already eroded and if you can use these masks to extract average signal for nuisance regression for task-based fMRI. I found that the noise components for ComCor noise ROIs are based on the eroded masks, but I am not sure if the masks provided by fmriprep in the anat folder are already eroded or not. Does anyone know this?

Kind regards,