Slice timing in afni_proc.py with multiple runs

[ryannct]

AFNI version info (afni -ver): 24.2.1

Hi AFNI experts!

I am preprocessing some task fMRI data that has 2 runs per session, per subject. I am aware that with afni_proc you can specify multiple runs with a wildcard * in your -dsets argument. Similarly, our stimulus timing files are in the format of the timing_tool.py, with each row/line being a separate run. If I understand correctly, the wildcard will read run-01 and run-02 in that order, and since they are in "alphabetical" order, the first row in the stimulus timing file will be applied to run-01, and the second row will be applied to run-02.

However, I am wondering what to do when you have multiple runs with regard to the slice timing files? Currently I have made separate slice timing files for each run. An example of one of my slice timing files:

0.66 0 0.44 0.055 0.495 0.11 0.55 0.165 0.605 0.275 0.715 0.33 0.77 0.385 0.825 0.22 0.66 0 0.44 0.055 0.495 0.11 0.55 0.165 0.605 0.275 0.715 0.33 0.77 0.385 0.825 0.22 0.66 0 0.44 0.055 0.495 0.11 0.55 0.165 0.605 0.275 0.715 0.33 0.77 0.385 0.825 0.22

I'm wondering, should I be setting up my slice timing file (as specified under the -tpattern option for -tshift_opts_ts) similarly to my stimulus timing files, with one row/line per run? Or is there another way that I will have to specify the correct slice timing file per run?

Below is my afni_proc.py code, if it helps.

## 3. Run afni_proc.py
apptainer exec -B /work $AFNI_path afni_proc.py								\
    -dsets 				sub-${sub_id}_ses-${ses_id}_task-${task_id}_run-*_bold.nii							\
    -copy_anat				${prewarp_path}/anatSS.sub-${sub_id}_ses-${ses_id}.nii		\
    -anat_has_skull			no								\
    -anat_follower 			anat_w_skull anat ${T1_path}					\
    -blocks 				tshift align tlrc volreg mask blur scale regress		\
    -radial_correlate_blocks 		tcat volreg							\
    -tcat_remove_first_trs 		0								\
    -tshift_opts_ts 			-tpattern @${st_path}						\
    -align_unifize_epi 			local								\
    -align_opts_aea 			-cost lpc+ZZ -giant_move -check_flip				\
    -tlrc_base 				${MNI_path}							\
    -tlrc_NL_warp											\
    -tlrc_NL_warped_dsets										\
           				${prewarp_path}/anatQQ.sub-${sub_id}_ses-${ses_id}.nii		\
           				${prewarp_path}/anatQQ.sub-${sub_id}_ses-${ses_id}.aff12.1D	\
           				${prewarp_path}/anatQQ.sub-${sub_id}_ses-${ses_id}_WARP.nii	\
    -volreg_align_to 			MIN_OUTLIER							\
    -volreg_align_e2a											\
    -volreg_tlrc_warp											\
    -volreg_compute_tsnr 		yes								\
    -mask_epi_anat 			yes								\
    -blur_size 				6.0								\
    -regress_stim_types 		AM1								\
    -regress_basis 			'dmBLOCK'							\
    -regress_stim_times 		sub-${sub_id}_ses-${ses_id}_task-${task_id}_beh/*							\
    -regress_stim_labels 		text shape alc nonalc motor					\
    -regress_local_times										\
    -regress_opts_3dD											\
    					-gltsym 'SYM: +alc -nonalc'					\
    					-glt_label 1 alc_vs_nonalc					\
    					-allzero_OK							\
    					-GOFORIT 6							\
    					-jobs $cpu_num							\
     -regress_motion_per_run										\
     -regress_censor_motion 		0.2								\
     -regress_censor_outliers 		0.05								\
     -regress_3dD_stop											\
     -regress_reml_exec											\
     -regress_compute_fitts										\
     -regress_make_ideal_sum 		sum_ideal.1D							\
     -regress_est_blur_epits										\
     -regress_est_blur_errts										\
     -regress_run_clustsim 		no								\
     -html_review_style 		pythonic							\
     -execute

Thanks so much!

Ryann

[rickr]

Hi Ryann,

Please feel free to check the proc script, but that -tpattern file should apply to all runs. Note that you can -copy_files $st_path to make a local one, if there is any preference to do so. I would personally do that, just to have the slice timing information locally, but it is a little thing.

• rick

[ryannct]

So I've looked at the .json files that I have for the different runs (within same person and session), and I'm a little worried because the slice timing in the .json files is just slightly different for some of the subjects (not all). These are the slice timing in the .json files of the same participant, same session, same tasks, run 01 vs run 02 (truncated):

vs.

This data was collected back in 2017 or 2018 by someone else so I'm not sure how that could have happened for the runs of the same task in the same session. Is there a way that I can try apply separate -tpattern files for each run?

[dglen]

Siemens has a 2.5ms/400Hz limitation in their DICOM files. I believe the precision of physiological recording devices determines the recorded timing. See this post:

Slice timing: General Electrics versus Siemens

[ryannct]

Thanks @dglen, that's very interesting. I've only worked with GE data before this so I wasn't aware that it would record it differently. Seems still weird to me that 2 runs of the same task in the same session, which I think were done back to back, would have different slice timings, and only for some subjects/sessions, not all.

In terms of my next steps -- do you think I could just use one slice time file for both runs (even if it's not an exact match for the slice timing of the other run)? Or will this really mess up my data (this is task data)?

Thanks!

[dglen]

That's probably okay. It looks like you have 18 slices, so each slice will be offset by 1/18, (0.0555555). The slice timing pattern appears to be alt+z2, with slice 1 (the secoond slice) acquired first. You can specify alt+z2 as the tpattern for 3dTshift to have 3dTshift calculate this too.

[ryannct]

Great, thanks so much! I really appreciate the help!