Project fmriprep-reproducibility

Dataset structure

• All inputs (i.e. building blocks from other sources) are located in inputs/.
• Outputs are available at outputs/.
• The code is located in fmriprep-reproducibility/.

Installation

The fmriprep-reproducibility rely heavily on Datalad, so make sure it is installed in your environment.

datalad install -r -R 1 git@github.com:SIMEXP/fmriprep-lts.git

When the installation finished, you can install the other software dependencies needed by our tool:

make install

Data

We carefully selected various datasets from openneuro to create a versatile dataset categorized by age, sex, and conditions. It also include different fieldmaps technics and non-parametric structural MR images types. There is another sub-dataset that consist of different singularity containers that we are using.

To install all the data:

make data

For a complete overview of all the used datasets, please read this. And for more information on the singularity containers used, check this section.

Usage

We provide an utility script for any slurm-compatible HPC that can help you spawn your fmriprep processes.

run.bash

--submit                whether or not to submit the slurm files
--slurm-script          select on which slurm script you want your experiments to be based on (default: fmriprep-slurm.bash)
--fmriprep-version      which fmriprep container version to use (default: 20.2.1)
--sampling              what sampling method to use between ieee or fuzzy (default: ieee)
-a|--account            slurm account
-m|--mail               mail for slurm notifications

The recommended way to test the workflow is:

./fmriprep-reproducibility/run.bash --slurm --account def-XXXX --mail XXX@mail.com

Reports

To generate html reproducibility figures of the functionnal pipeline, you can run:

make report

There are multiple arguments available, for example to list all available arguments call:

make report ARGS="--help"

Execution environment

Usage

Singularity images are available in sub-dataset singularity-images stored on OSF at https://osf.io/rbu92. The sub-dataset can be installed as follows:

datalad install envs/singularity-images

The Singularity image can be downloaded as follows:

datalad get envs/singularity-images/fmriprep-lts*

And it can be used as follows:

singularity run ./fmriprep-lts.sif <args>

Fuzzy mode

fmriprep can be executed in "fuzzy" mode as follows:

singularity run ./fmriprep-lts-fuzzy-VERSION.sif <args>

This mode simulates machine error by introducing random perturbations in the results of mathematical functions. It is used to build a reference variability map in the tests.

Building images

The Singularity image is built from the official fmriprep Docker image. Building the image requires local installations of:

• Docker
• Singularity >= 3.5

The image is built by running build.sh from the base directory of this repo.

Once they are built, images can be pushed to OSF as follows:

OSF_TOKEN=<your_personal_token> datalad push --to github