Hi both,
Thanks for this discussion, I am thinking about adopting the procedures you outlined here. I am just having some trouble thinking about how you can integrate confounds from T1 space and the AROMA noise IC’s in MNI space - are the WM/CSF confounds in T1 space still valid or do they need to recalculated.
Ideally, I would like to extract ROI timecourses (aparcaseg) in T1 space that have been denoised using ICA-AROMA, WM, CSF, high pass filter and a linear detrend. But I am not sure if it is valid to use the AROMA noise components in data from T1 space. Do you have any thoughts on that?
Perhaps a similar approach to this post: Preprocessing for atlas based connectivity matrices
Aslo @jdkent concerning your tests on confound regressions (https://github.com/poldracklab/fmriprep/issues/817): was it slightly better to regress the WM and CSF signals in T1 space from the smoothAROMAnonaggr_bold.nii.gz, as opposed to recalculating the regressors after ICA-AROMA?
Many thanks,
Joff