Questions regarding noise reduction in AFNI Proc

AFNI Message Board
1

Hi,

I have a question about my AFNI afni_proc.py script (see below). My main goal is to remove as much noise as possible using AFNI alone, without using Freesurfer.

Of course, what constitutes "noise" versus "signal" can be subjective, and it particularly depends on the type of measurements applied later in the analysis (e.g., in Python after AFNI preprocessing).

I have a few specific questions:

  1. Is it appropriate to use both -regress_ROI CSFe and -regress_ROI_PC Svent 3 together? I understand that the former performs regression and the latter applies PCA, but is combining them potentially problematic?

  2. Are there any other recommended options I could include in my script to further reduce noise that I may have overlooked? I'm planning to compute measurements like Sample Entropy and spectral analyses on the preprocessed time-series. That’s why I’m using a "resting-state" preprocessing pipeline, even though the run is actually a movie-watching run.

  3. Are there any other potential flaws ("mistakes" or "better don't do that") in my script?

afni_proc.py \
-subj_id Sub01_${run} \
-out_dir $dir_out/Results \
-dsets $dir_func/${run}_Sub01+orig \
-blocks despike tshift align tlrc volreg mask regress \
-copy_anat $dir_sswarper/anatSS.Sub01.nii \
-anat_has_skull no \
-align_unifize_epi local \
-align_opts_aea -cost lpc+ZZ \
				-giant_move \
				-check_flip \
-volreg_align_e2a \
-volreg_align_to MIN_OUTLIER \
-volreg_tlrc_warp -tlrc_base MNI152_2009_template_SSW.nii.gz \
-tlrc_NL_warp \
-tlrc_NL_warped_dsets \
	$dir_sswarper/anatQQ.Sub01.nii \
	$dir_sswarper/anatQQ.Sub01.aff12.1D \
	$dir_sswarper/anatQQ.Sub01_WARP.nii \
-volreg_post_vr_allin yes \
-volreg_pvra_base_index MIN_OUTLIER \
-mask_segment_anat yes \
-mask_segment_erode yes \
-mask_intersect Svent CSFe WMe \
-regress_polort 2 \
-regress_anaticor \
-regress_ROI CSFe \
-regress_ROI_PC Svent 3 \
-regress_ROI_PC WMe 3 \
-regress_ROI_PC_per_run Svent WMe \
-regress_apply_mot_types demean deriv \
-regress_motion_per_run \
-regress_censor_motion 0.3 \
-regress_censor_outliers 0.05 \
-regress_skip_first_outliers 5 \
-html_review_style pythonic \
-execute