apaper_afni_proc

Scripts to accompany FMRI processing examples using AFNI's afni_proc.py, for the accompanying paper:

• Reynolds RC, Glen DR, Chen G, Saad ZS, Cox RW, Taylor PA (2024). Processing, evaluating and understanding FMRI data with afni_proc.py.
  arXiv:2406.05248 [q-bio.NC] https://arxiv.org/abs/2406.05248

Full sets of the data and processing scripts (which themselves are copies of script directories in the present git repository, from the time of publication) are available for download, as described below. There are two downloadable demos provided: one for resting state FMRI (acquired with multiple echos, and including processing for both single- and multi-echo FMRI) and one for task-based FMRI.

How to use the scripts (general organization and guidance)

Desktop and HPC scripts

In each demo download, a pair of script directories is provided:

• scripts_*_biowulf : scripts set up for batch processing on a high performance cluster that uses Slurm as its workload manager (specifically, swarm to submit multiple jobs simultaneously), which is called Biowulf at NIH where the processing was originally performed.
• scripts_*_desktop : scripts set up for batch processing on a standard desktop running Linux or macOS.

In each download, only one subject's data is provided, but the scripts are setup to process a group of subjects with unprocessed data organized in the same way. That means there is more "scripty" stuff to read, rather than just FMRI processing commands, but we hope it is useful.

Running scripts

In each case, scripts are organized to run in the following way, partitioning the "what needs to be done per subject" from the "managing the looping over a group of subjects". Therefore, each stage of processing is controlled by a pair of scripts, the do_*.tcsh and run_*.tcsh script, respectively:

• do_*.tcsh : do one stage of processing (like nonlinear alignment, running FreeSurfer, running afni_proc.py, etc.) on one dataset, whose subject ID (and possibly session ID) are input at runtime.
• run_*.tcsh : manage having one or more datasets to process, such as by looping over all subject (and possibly session) IDs in a directory, and either setup a swarm script to run on an HPC or start processing in series on a desktop.

The user primarily executes the "run" script, which itself calls the associated "do" script one or more times. Each "do-run" pair produces one new data directory containing a directory per subject of the output results of that processing stage.

Script and data tree naming

The script names contain a 2-digit number near the beginning, so that a simple ls in the directory lists them in the approximate order of expected execution. That is, run_03*.tcsh comes before run_22*.tcsh. There are gaps in the numbering, to leave room for other stages to be inserted when adapting them. Also, sometimes the numbering just separates stages that would be run in parallel; for example, each afni_proc.py example is independent, and these are simply each of the run_2*.tcsh and run_3*.tcsh scripts here.

There is a simple string label associated with each number, that remains constant for both the do_*.tcsh and run_*.tcsh scripts, as well as the output data directory. Thus, do_03_timing.tcsh is paired with run_03_timing.tcsh, which produces data_03_timing as the output data tree.

Task FMRI demo (Ex. 2, and supplements)

Download

The accompanying data can be downloaded+unpacked via:

curl -O https://afni.nimh.nih.gov/pub/dist/tgz/demo_apaper_afni_proc_task.tgz
tar -xf demo_apaper_afni_proc_task.tgz

Raw data contents

The raw/unprocessed data tree is located in data_00_basic/, with the following BIDS file structure (though afni_proc.py does not require any particular file structure to run processing):

data_00_basic/
|-- dataset_description.json
|-- participants.tsv
|-- README
`-- sub-10506
    |-- anat
    |   |-- sub-10506_T1w.json
    |   `-- sub-10506_T1w.nii.gz
    `-- func
        |-- sub-10506_task-pamenc_bold.json
        |-- sub-10506_task-pamenc_bold.nii.gz
        `-- sub-10506_task-pamenc_events.tsv

Scripts already run

The following processing script pairs exist in the scripts*/ directory, and have already been run so their data directories exist in the distributed demo:

• *03_timing* : run timing_tool.py to generate timing files from the stimulus events TSV
• *12_fs* : run FreeSurfer's recon-all and AFNI's @SUMA_Make_Spec_FS on the raw anatomical dataset, to estimate surface meshes and then create standardized mesh versions; also creates anatomical parcellations.
• *13_ssw* : run sswarper2 on the raw anatomical dataset, to skullstrip (ss) it and to estimate nonlinear alignment (warping) to standard space.

NB: a subset of output datasets (trimmed to what is useful for further steps) might be distributed in some cases, to reduce the download size of the demo.

Scripts to be run

The following processing script pairs exist in the scripts*/ directory, for the user to run:

• *20_ap_simple* : run a "simple" afni_proc.py command on the raw data, that requires essentially no options and does only affine alignment to template space for quicker processing, but generates a very useful set of outputs and QC HTML for investigating the data quickly.
• *22_ap_ex2_task* : run the Ex. 2 afni_proc.py command for task-based FMRI processing. This includes using nonlinear alignment, stimulus timing, and more.
• *40_ap_ex10_task_bder* : run the (supplementary) Ex. 10 afni_proc.py command for task-based processing. This is a variation of Ex. 2 that includes an output directory in BIDS-Derivative format.

Supplementary scripts

• pack*.tcsh* : several of the above scripts run programs that create QC output; these scripts pack up things like the QC images or QC HTML directories from a given output across all subjects and create a compressed tarball, which can be more easily moved around.

Rest FMRI Demo (Ex. 1, 3, 4 and supplements)

The accompanying data can be downloaded+unpacked via:

curl -O https://afni.nimh.nih.gov/pub/dist/tgz/demo_apaper_afni_proc_rest.tgz
tar -xf demo_apaper_afni_proc_rest.tgz

The raw data tree is located in data_00_basic/, ***

data_00_basic/
`-- sub-005
    `-- ses-01
        |-- anat
        |   `-- sub-005_ses-01_mprage_run-1_T1w.nii.gz
        `-- func
            |-- sub-005_ses-01_acq-blip_dir-match_run-1_bold.nii.gz
            |-- sub-005_ses-01_acq-blip_dir-opp_run-1_bold.nii.gz
            |-- sub-005_ses-01_task-rest_run-1_echo-1_bold.nii.gz
            |-- sub-005_ses-01_task-rest_run-1_echo-2_bold.nii.gz
            |-- sub-005_ses-01_task-rest_run-1_echo-3_bold.nii.gz
            |-- sub-005_ses-01_task-rest_run-1_physio-ECG.txt
            `-- sub-005_ses-01_task-rest_run-1_physio-Resp.txt

Scripts already run

The following processing script pairs exist in the scripts*/ directory, and have already been run so their data directories exist in the distributed demo:

• *12_fs* : run FreeSurfer's recon-all and AFNI's @SUMA_Make_Spec_FS on the raw anatomical dataset, to estimate surface meshes and then create standardized mesh versions; also creates anatomical parcellations.
• *13_ssw* : run sswarper2 on the raw anatomical dataset, to skullstrip (ss) it and to estimate nonlinear alignment (warping) to standard space.
• *14_physio* : run physio_calc.py on the physiological (cardiac and respiratory) time series, to estimate physio regressors for the FMRI processing.

NB: a subset of output datasets (trimmed to what is useful for further steps) might be distributed in some cases, to reduce the download size of the demo.

Scripts to be run

The following processing script pairs exist in the scripts*/ directory, for the user to run:

• *20_ap_simple* : run a "simple" afni_proc.py command on the raw data, that requires essentially no options and does only affine alignment to template space for quicker processing, but generates a very useful set of outputs and QC HTML for investigating the data quickly.
• *21_ap_ex1_align* : run Ex. 1's afni_proc.py command that basically just does the alignment part of FMRI processing, which appropriate concatenation; includes blip up/down (B0 inhomogeneity) distortion correction.
• *23_ap_ex3_vol* : run Ex. 3's afni_proc.py command, which is a fairly standard resting state processing case for voxelwise analysis; includes physio regressors, GM ROIs from FreeSurfer, and user-specified ROI imports for TSNR checks.
• *24_ap_ex4_mesurf* : run Ex. 4's afni_proc.py command, which processes multi-echo (ME) FMRI data and performs surface-based analysis; the tedana-MEICA is used to combine echos; blip up/down correction is also performed.
• *35_ap_ex5_vol* : run (supplementar) Ex. 5's afni_proc.py command, which is a variation of Ex. 3 that includes bandpassing; see the draft for a discussion of this topic.
• *36_ap_ex6_vol* : run (supplementar) Ex. 6's afni_proc.py command, which is a variation of Ex. 3 that includes local ANATICOR and ventricle regressors to the processing; also includes blurring to FWHM.
• *37_ap_ex7_roi* : run (supplementar) Ex. 7's afni_proc.py command, which is a variation of Ex. 3 that performs processing for an ROI-based study; primarily, this means not including blurring.
• *38_ap_ex8_mesurf* : run (supplementar) Ex. 8's afni_proc.py command, which is a variation of Ex. 4 that uses OC to combine ME-FMRI echos.
• *39_ap_ex9_mevol* : run (supplementar) Ex. 9's afni_proc.py command, which is a variation of Ex. 3 that processes ME-FMRI in the volume.

Supplementary scripts

• pack*.tcsh* : several of the above scripts run programs that create QC output; these scripts pack up things like the QC images or QC HTML directories from a given output across all subjects and create a compressed tarball, which can be more easily moved around.