I’ve noticed that some resting-state preprocessing pipelines include regressing out signal from the white matter and CSF. However, I’m wondering if this is a necessary step in FATCATs 3dROIMaker, if we select the “trim_off_wm” ? It seems to me that in a sense, that is regressing out the signal.
What are your thoughts?
Typically, we would use something like a FreeSurfer-based anatomical parcellation to determine the location of the WM and CSF (or, more specifically for the latter, the ventricles); these would be provided to afni_proc.py and eroded therein before using them to map out tissue-based regressor locations. The erosion is important because we do not want them to overlap GM at all.
So, if you have good WM and venticle maps, you could use those, sure.
I’m a bit curious where you are intending to get your WM maps, if you are mentioning 3dROIMaker? Are you meaning to use DTI-defined WM (say, where FA>0.2) to create the tissue map? I would probably run FreeSurfer instead, as there might me some high-FA locations that happen to overlap the GM.
Yes I was going to use DTI-defined white matter to create the tissue type. Would that be okay?
I can also look into freesurfer, but I was just wondering if it would make sense to use the 3dROIMaker, where I can filter out CSF and white matter. I don’t see how much different that would be.
I think FreeSurfer would be better; eroding the FA>0.2 mask of the DTI-definied WM would probably be pretty good, but the FreeSurfer one should be more reliable. Here is a brief tutorial on running+using FreeSurfer with AFNI:
… including converting its results to usable NIFTI and GIFTI files afterwards, with @SUMA_Make_Spec_FS.
To check either set of results (which should be done), you would want to see that the WM and CSF maps do not overlap any GM.