Remarks on FMRI processing, with example scripts in afni_proc.py
Poster No:
1570
Submission Type:
Abstract Submission
Authors:
Paul Taylor1, Daniel Glen2, Richard Reynolds3
Institutions:
1National Institute of Mental Health, Bethesda, MD, 2National Institute of Mental Health (NIMH), NIH, Bethesda, MD, 3NIMH, Bethesda, MD
First Author:
Co-Author(s):
Introduction:
FMRI processing is complicated. It relies on various computational procedures, including alignment, "data cleaning" (e.g. despiking and censoring), time series analysis and statistical modeling. Researchers perform many different types of studies, each with a particular acquisition and modeling design. Making pipeline steps consistent with study goal aims is an important procedure.
Here, we discuss some ways of matching processing choices with various study designs and data properties. We describe some general tips and considerations, which may apply across all software tools. We also present a set of example scripts that implement a number of related processing procedures (such as nonlinear alignment, timing file creation and physio regressor estimation) in the open source, widely used AFNI toolbox [1] and its FMRI pipeline generation tool, afni_proc.py [2]. These include both task-based and resting state examples, with volumetric and surface processing.
Here, we discuss some ways of matching processing choices with various study designs and data properties. We describe some general tips and considerations, which may apply across all software tools. We also present a set of example scripts that implement a number of related processing procedures (such as nonlinear alignment, timing file creation and physio regressor estimation) in the open source, widely used AFNI toolbox [1] and its FMRI pipeline generation tool, afni_proc.py [2]. These include both task-based and resting state examples, with volumetric and surface processing.
Methods:
We discuss some frequently asked questions about FMRI processing options. We also provide a publicly available set of related example scripts here: https://github.com/afni/apaper_afni_proc/. Each script packet is commented and constructed to apply across a group of subjects (which can be in BIDS trees). There are dual sets of scripts: one for desktop execution, and one for swarming on HPC cluster systems. When executed, afni_proc.py creates a fully commented script of the pipeline, to further help researchers understand the processing choices and implementation.
Results:
Should I blur/smooth the FMRI data, and by how much? It is common to blur in voxelwise studies (in afni_proc.py, via the "blur" block). For single echo FMRI, one might blur 1.5-2 times the minimum voxel size. Multi-echo FMRI has higher TSNR, so one might blur just slightly above voxel dimension. When performing ROI-based studies, blurring should not be applied (and one should not include the "blur" block), so that ROI averages are not corrupted from outside the ROIs.
How can I reduce EPI distortion? In addition to optimizing scanner parameters, one can acquire phase images (field maps) or an opposite phase-encoded EPI [3]; each can be integrated directly with afni_proc.py. Neither can fully remove distortion, but each helps and adds negligible scan time. In practice, using opposite phase-encoded EPI may have slight advantages in most software (e.g., [4]). Acquiring 5-10 reverse encoded volumes reduces the chance of subject motion ruining the set, and typically takes only 10-20 s total. One can add the phase pairs to afni_proc.py with "-blip_forward_dset" and "-blip_reverse_dset". Viewing both the raw EPI and results of EPI-anatomical alignment is a useful way to judge distortions (Fig 1), as shown in afni_proc.py's quality control HTML [5].
Are there any downsides to bandpassing resting state data between 0.01-0.1 Hz? **Yes**. First, bandpassing should typically be applied as part of the regression step (not separate Fourier), to avoid spectral leakage and other artifacts [2,6]. Also, bandpassing is statistically expensive and removes a large fraction of degrees of freedom (DFs): 60% when TR=2s, 80% when TR=1s. This may create mathematical issues when censoring is applied [7]; afni_proc.py checks DFs carefully to warn users. There has also been shown to be meaningful signal patterns above 0.1 Hz, so useful features may be removed from the data [8,9].
How can I reduce EPI distortion? In addition to optimizing scanner parameters, one can acquire phase images (field maps) or an opposite phase-encoded EPI [3]; each can be integrated directly with afni_proc.py. Neither can fully remove distortion, but each helps and adds negligible scan time. In practice, using opposite phase-encoded EPI may have slight advantages in most software (e.g., [4]). Acquiring 5-10 reverse encoded volumes reduces the chance of subject motion ruining the set, and typically takes only 10-20 s total. One can add the phase pairs to afni_proc.py with "-blip_forward_dset" and "-blip_reverse_dset". Viewing both the raw EPI and results of EPI-anatomical alignment is a useful way to judge distortions (Fig 1), as shown in afni_proc.py's quality control HTML [5].
Are there any downsides to bandpassing resting state data between 0.01-0.1 Hz? **Yes**. First, bandpassing should typically be applied as part of the regression step (not separate Fourier), to avoid spectral leakage and other artifacts [2,6]. Also, bandpassing is statistically expensive and removes a large fraction of degrees of freedom (DFs): 60% when TR=2s, 80% when TR=1s. This may create mathematical issues when censoring is applied [7]; afni_proc.py checks DFs carefully to warn users. There has also been shown to be meaningful signal patterns above 0.1 Hz, so useful features may be removed from the data [8,9].
Conclusions:
afni_proc.py allows users to control a large number of details about the processing, and it is helpful to know what considerations apply for a study aim. We provide some of these here, as well as a set of starter examples. An afni_proc.py command is easily and openly sharable; researchers can add their own comments. These discussion points may help users of any FMRI processing software tool, since similar considerations apply regardless of the tool. Understanding processing details may also likely facilitate cross-software reproducibility.
Modeling and Analysis Methods:
Methods Development 1
Other Methods 2
Keywords:
Data analysis
FUNCTIONAL MRI
Workflows
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Are you Internal Review Board (IRB) certified? Please note: Failure to have IRB, if applicable will lead to automatic rejection of abstract.
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
[2] Reynolds RC, Glen DR, Chen G, Saad ZS, Cox RW, Taylor PA (2024). Processing, evaluating and understanding FMRI data with afni_proc.py. Imaging Neuroscience 2:1-52.
https://doi.org/10.1162/imag_a_00347
[3] Chang H, Fitzpatrick JM (1992). A technique for accurate magnetic resonance imaging in the presence of field inhomogeneities. IEEE Trans. Med. Imaging, 11:319-329
[4] Roopchansingh V, French JJ, Nielson D, Reynolds R, Glen D, D’Souza P, Taylor P, Cox R, Thurm A (2020). EPI Distortion Correction is Easy and Useful, and You Should Use It: A case study with toddler data. bioRxiv 2020.09.28.306787; doi: https://doi.org/10.1101/2020.09.28.306787
[5] Taylor PA, Glen DR, Chen G, Cox RW, Hanayik T, Rorden C, Nielson DM, Rajendra JK, Reynolds RC (2024). A Set of FMRI Quality Control Tools in AFNI: Systematic, in-depth and interactive QC with afni_proc.py and more. Imaging Neuroscience 2: 1–39. doi: 10.1162/imag_a_00246
[6] Hallquist MN, Hwang K, Luna B (2013). The nuisance of nuisance regression: spectral misspecification in a common approach to resting-state fMRI preprocessing reintroduces noise and obscures functional connectivity. Neuroimage. 82:208–225. doi:10.1016/j.neuroimage.2013.05.116
[7] Caballero-Gaudes C, Reynolds RC (2017). Methods for cleaning the BOLD fMRI signal. Neuroimage 154:128-149. doi: 10.1016/j.neuroimage.2016.12.018
[8] Gohel SR, Biswal BB (2015). Functional integration between brain regions at rest occurs in multiple-frequency bands. Brain Connectivity. 5(1):23-34.
[9] Shirer WR, Jiang H, Price CM, Ng B, Greicius MD (2015). Optimization of rs-fMRI pre-processing for enhanced signal-noise separation, test-retest reliability, and group discrimination. Neuroimage. 117:67-79.