afni_proc script and QC

AFNI Message Board
1

Hello,

I created my afni.proc script based on examples I saw on the afni website. I have ran this script on my participant’s and it runs successfully. I have checked the QC html and the outputs seem good. I have two questions based on my script and QC I am planning on doing:

(1) I was wondering whether my script looks okay and whether there is anything integral missing or anything else I should add? My script is below.
(2) In terms of QC, I am going through the ss_review_driver and QC_xxx/index.html -is there anything else you recommend that I should evaluate to complete the QC?
(3) Within the ss_review_basic, I have been reviewing my participant’s avg censored motion, max censored displacement and censor fraction, for high numbers. I was wondering whether there are any threshold levels I should use to potentially remove participants?

Thank you very much,
Tamara

afni_proc.py -subj_id 1422 -script proc.June2_2021 -scr_overwrite
-blocks tshift align tlrc volreg blur mask scale regress -copy_anat
/imaging/1422/1422.anat+orig -dsets
/imaging/1422/1422.1+orig.HEAD
/imaging/1422/1422.2+orig.HEAD
/imaging/1422/1422.3+orig.HEAD
/imaging/1422/1422.4+orig.HEAD
/imaging/1422/1422.5+orig.HEAD
/imaging/1422/1422.6+orig.HEAD
-tcat_remove_first_trs 0 -tlrc_base MNI152_T1_2009c+tlrc -tlrc_NL_warp
-volreg_align_to MIN_OUTLIER -volreg_align_e2a -volreg_tlrc_warp -blur_size
6.0 -regress_stim_times
/imaging/1422/Dfeel_Bad_NoResp.txt
/imaging/1422/Dfeel_Bad_Resp.txt
/imaging/Dfeel_Happy_NoResp.txt
/imaging/Dfeel_Happy_Resp.txt
/imaging/Feel_Bad_NoResp.txt
/imaging/Feel_Bad_Resp.txt
/imaging/Feel_Happy_NoResp.txt
/imaging/Feel_Happ_Resp.txt
/imaging/Nat_neg_NoResp.txt
/imaging/Nat_neg_Resp.txt
/imaging/Nat_post_NoResp.txt
/imaging/1422/Nat_post_Resp.txt
/imaging/1422/Rating.txt
/imaging/1422/new.Instruct.txt
-regress_stim_labels Dfeel_Bad_NoResp Dfeel_Bad_Resp Dfeel_Happy_NoResp
Dfeel_Happy_Resp Feel_Bad_NoResp Feel_Bad_Resp Feel_Happy_NoResp
Feel_Happy_Resp Nat_neg_NoResp Nat_neg_Resp Nat_post_NoResp
Nat_post_Resp Rating Instruct
-regress_basis dmBLOCK -regress_stim_types AM1 -regress_censor_motion
1.0 -regress_censor_outliers 0.1 -regress_apply_mot_types demean deriv
-regress_motion_per_run -regress_make_ideal_sum sum_ideal.1D
-regress_est_blur_epits -regress_est_blur_errts -regress_run_clustsim no
-regress_opts_3dD -GOFORIT 15
-execute

2

Hi Tamara,

(1) That looks reasonable, and seems similar to “Example 6b” from the help. But it might be worth you considering some of the differences.

Add this to your command:

-compare_opts ‘example 6b’

(and possibly -verb 2), and see what it says.

One example is that we suggest running @SSwarper on each subject before running afni_proc.py, and then passing the warp info as in Example 6b.

For QC, it is nice to include the -radial_correlate_blocks option.
Example 6 also uses 3dREMLfit for the regression, which is somewhat preferable to the OLSQ in 3dDeconvolve.

Also, -html_review_style pythonic will give nicer plots (if you have matplotlib) in the QC report.

(2) With the QC specified, note that the final output from the analysis should be a suggestion to open a QC HTML report. That report is a nice QC review and should be the minimum review for each subjects. For the first couple/few subjects, it is suggested to review the data at the level of the bootcamp class “Start to Finish”. We have videos on that (17 videos covering 6 hours, it is detailed).

(3) Those are reasonable parameters to focus on, but the levels depend on the subjects. Note that for healthy adults, those censor limits are high. But some subjects are much more prone to motion.

Motion (damaging estimates of the betas) can subsequently hurt the group results. So if you end up dropping 5-20% of the subjects, it might actually help. There are no universal numbers because of how they vary across subject groups.

  • rick