fMRIPrep

fMRIPrep was developed by Russ Poldrack and his team at Stanford University. It is an open-source functional magnetic resonance imaging (fMRI) pre-processing pipeline. This page offers a brief introduction to the software with links to useful resources.

This application was developed to perform minimal, standardized pre-processing steps with a combination of tools from state-of-the-art software packages (e.g., FSL, ANTs, FreeSurfer, and AFNI). Default pre-processing consists of motion correction, field unwarping, normalization, bias field correction, and brain extraction.

Usage

fMRIPrep requires at least 2 inputs: a T1w image and a BOLD series. Input data must be in BIDS format. Please see the BIDS page for more details.

fMRIPrep can be used and installed through 2 methods: a containerized image (Docker or Singularity/Apptainer) and Manually Prepared Environment (Python 3.10+). A containerized method is strongly recommended as it minimizes the need for the tedious installation of it's many dependencies. An Apptainer is available for those with access to the takim2 host of the PMACS LPC at /project/singularity_images/fmriprep_latest.simg. Below is a brief example using this image:

apptainer run --cleanenv \
    # bind project directory \
    -B <project-directory>:<project-directory> \
    # bind input directory \
    -B <input-directory>:<input-directory> \
    # bind output directory \
    -B <output-directory>:<output-directory> \
    # bind working directory \
    -B <working-directory>:<working-directory> \
    # bind scratch directory \
    -B <scratch-directory>:/scratch \
    # fMRIPrep Apptainer image \
    /project/singularity_images/fmriprep_latest.simg \
    # define input and output directories \
    <input-directory> <output-directory> \
    # process individual participant(s) \
    participant \
    # select participant(s) to process \
    --participant-label sub-<subject-ID> \
    # max threads used across workflow \
    --nthreads <max-threads> \
    # max threads available for a single process \
    --omp-nthreads <max-threads-per-process> \
    # memory limit in megabytes \
    --mem-mb <memory-limit> \
    # reduce memory usage \
    --low-mem \
    # FreeSurfer license \
    --fs-license-file <project-directory>/license.txt \
    # specify working directory \
    -w <working-directory>

More details and other optional arguments can be found here.

Note that a FreeSurfer license is required for use and defined with --fs-license-file. You can register for free here.

Output

After an fMRIPrep run, an html report for each subject is created for visual inspection and quality control. Pre-processed data is saved in <output-directory>/sub-<subject-ID>, which can be used for further processing or subsequent analyses.

Resources

For other useful resources and tutorials, I strongly recommend checking out Andy's Brain Book! Neurostars is an open-access question and answer forum with a large knowledge base.