Hi there, a reviewer recently suggested using MELODIC ICA to remove noise from my data (task-based fmri) - and then to use those noise regressors in my GLM. I've used AFNI to preprocess my data (see details below) and am using 3dDeconvolve to run individual regression analyses.
I'm wondering if there are any thoughts on using MELODIC output within the AFNI pipeline? From what I understood, I was controlling for noise and motion using my regressors of no interest, so would it be redundant to include additional noise regressors? Since AFNI doesn't have ICA features, I was thinking this was controlled for in other ways for task data? I am hesitant to combine different software pipelines - so I would appreciate any input/thoughts. Thank you in advance!
AFNI preprocessing includes: 3dSkullStrip, 3dDespike, 3dToutcount, align_epi_anat.py, with 3dAllineate, adwarp, @auto_tlrc, spatial smoothing (4-mm FWHM, 3dmerge), and scaling (3dTstat, 3dcalc)
my GLM code:
3dDeconvolve -input subject_run_scale+tlrc.HEAD
-stim_times 1 /extdata/project/scripts/pro_mixed.1D 'GAM' -stim_label 1 'pro_event'
-stim_times 2 /extdata/project/scripts/anti_mixed.1D 'GAM' -stim_label 2 'anti_event'
-stim_times 3 /extdata/project/scripts/mixed_block.1D 'BLOCK(48,1)' -stim_label 3 'block'
-stim_file 4 subject_motion_demean_run.1D'' -stim_base 4 -stim_label 4 Roll
-stim_file 5 subject_motion_demean_run.1D'' -stim_base 5 -stim_label 5 Pitch
-stim_file 6 subject_motion_demean_run.1D'' -stim_base 6 -stim_label 6 Yaw
-fout -rout -tout -bout -nofull_first