Defining standard and nonstandard spaces where data will be resampled

The command line interface of fMRIPrep allows resampling the preprocessed data onto other output spaces. That is achieved using the --output-spaces argument, where standard and nonstandard spaces can be inserted.

Standard spaces

When using fMRIPrep in a workflow that will investigate effects that span across analytical groupings, neuroimagers typically resample their data on to a standard, stereotactic coordinate system. The most extended standard space for fMRI analyses is generally referred to MNI. For instance, to instruct fMRIPrep to use the MNI template brain distributed with FSL as coordinate reference the option will read as follows: --output-spaces MNI152NLin6Asym. By default, fMRIPrep uses MNI152NLin2009cAsym as spatial-standardization reference. Valid template identifiers (MNI152NLin6Asym, MNI152NLin2009cAsym, etc.) come from the TemplateFlow project.

Therefore, fMRIPrep will run nonlinear registration processes against the template T1w image corresponding to all the standard spaces supplied with the argument --output-spaces. By default, fMRIPrep will resample the preprocessed data on those spaces (labeling the corresponding outputs with the space-<template-identifier> BIDS entity) but keeping the original resolution of the BOLD data to produce smaller files, more consistent with the original data gridding. However, many users will be interested in utilizing a coarse gridding (typically 2mm isotropic) of the target template. Such a behavior can be achieved applying modifiers to the template identifier, separated by a : character. For instance, --output-spaces MNI152NLin6Asym:res-2 MNI152NLin2009cAsym will generate preprocessed BOLD 4D files on two standard spaces (MNI152NLin6Asym, and MNI152NLin2009cAsym) with the template’s 2mm isotropic resolution for the data on MNI152NLin6Asym space and the original BOLD resolution (say, e.g., 2x2x2.5 [mm]) for the case of MNI152NLin2009cAsym.

Other possible modifiers are, for instance, the cohort selector. Although currently there is no template in TemplateFlow with several cohorts, very soon we will integrate pediatric templates, for which cohort will function to select the appropriate age range. Therefore, in upcoming versions of fMRIPrep, it will be possible to run it with --output-spaces MNIPediatricAsym:res-2:cohort-2 where cohort-2 would select the template instance for the, say, 24-48 months old range.

When specifying surface spaces (e.g., fsaverage), the legacy identifiers from FreeSurfer will be supported (e.g., fsaverage5) although the use of the density modifier would be preferred (i.e., fsaverage:den-10k for fsaverage5). When selecting fsLR surface space without a density label, fsLR:den-32k will be used.

Custom standard spaces

Although the functionality is not available yet, the interface of the --output-spaces permits providing paths to custom templates that follow TemplateFlow’s naming conventions (e.g., /path/to/custom/templates/tpl-MyCustom:res-2). Following the example, at least the following files must be found under under /path/to/custom/templates/tpl-MyCustom:

tpl-MyCustom/
    template_description.json
    tpl-MyCustom_res-1_T1w.nii.gz
    tpl-MyCustom_res-1_desc-brain_mask.nii.gz
    tpl-MyCustom_res-2_T1w.nii.gz
    tpl-MyCustom_res-2_desc-brain_mask.nii.gz

Although a more comprehensive coverage of standard files would be advised.

Nonstandard spaces

Additionally, --output-spaces accepts identifiers of spatial references that do not generate standardized coordinate spaces:

  • T1w or anat: data are resampled into the individual’s anatomical reference generated with the T1w and T2w images available within the BIDS structure.

  • fsnative: similarly to the anat space for volumetric references, including the fsnative space will instruct fMRIPrep to sample the original BOLD data onto FreeSurfer’s reconstructed surfaces for this individual.

  • func, bold, run, boldref or sbref can be used to generate BOLD data in their original grid, after slice-timing, head-motion, and susceptibility-distortion corrections. These keywords are experimental, and expected to change because additional nonstandard spaces are currently being discussed here.

Modifiers are not allowed when providing nonstandard spaces.

Preprocessing blocks depending on standard templates

Some modules of the pipeline (e.g., the ICA-AROMA denoising, the generation of HCP compatible grayordinates files, or the fieldmap-less distortion correction) operate in specific template spaces. When selecting those modules to be included (using any of the following flags: --use-aroma, --cifti-outputs, --use-syn-sdc) will modify the list of output spaces to include the space identifiers they require, should the identifier not be found within the --output-spaces list already. In other words, running fMRIPrep with --output-spaces MNI152NLin6Asym:res-2 --use-syn-sdc will expand the list of output spaces to be MNI152NLin6Asym:res-2 MNI152NLin2009cAsym.

TemplateFlow

Group inference and reporting of neuroimaging studies require that individual’s features are spatially aligned into a common frame where their location can be called standard. To that end, a multiplicity of brain templates with anatomical annotations (i.e., atlases) have been published. However, a centralized resource that allows programmatic access to templates was lacking. TemplateFlow is a modular, version-controlled resource that allows researchers to use templates “off-the-shelf” and share new ones.

In addition to the repository from which neuroimaging templates are redistributed, TemplateFlow also comprehends a Python client tool to access them programmatically when used as a library by other software, or interactively by humans. Therefore TemplateFlow is the software module that allows fMRIPrep to flexibly change, and dynamically pull down, new standardized template information.

How do you use TemplateFlow in the absence of access to the Internet?. This is a fairly common situation in HPCs systems, where the so-called login nodes have access to the Internet but compute nodes are isolated, or in PC/laptop enviroments if you are travelling. TemplateFlow will require Internet access the first time it receives a query for a template resource that has not been previously accessed. If you know what are the templates you are planning to use, you could prefetch them using the Python client. To do so, follow the next steps.

  1. By default, a mirror of TemplateFlow to store the resources will be created in $HOME/.cache/templateflow. You can modify such a configuration with the TEMPLATEFLOW_HOME environment variable, e.g.:

    $ export TEMPLATEFLOW_HOME=$HOME/.templateflow
    
  2. Install the client within your favorite Python 3 environment (this can be done in your login-node, or in a host with Internet access, without need for Docker/Singularity):

    $ python -m pip install -U templateflow
    
  3. Use the get() utility of the client to pull down all the templates you’ll want to use. For example:

    $ python -c "from templateflow.api import get; get(['MNI152NLin2009cAsym', 'MNI152NLin6Asym', 'OASIS30ANTs', 'MNIPediatricAsym', 'MNIInfant'])"
    

After pulling down the resources you’ll need, you will just need to make sure your runtime environment is able to access the filesystem, at the location of your TemplateFlow home directory. If you are a Singularity user, please check out TemplateFlow and Singularity.