Fmriprep is an fMRI data preprocessing pipeline that produces an output that can be submitted to a variety of group-level analyses, including task-based and resting-state fMRI, graph theoretical metrics, and surface or volume-based statistics. Some of the preprocessing steps performed by fmriprep include coregistration, normalization, unwarping, noise component extraction, segmentation, and skullstripping.

References:

• Esteban O, Markiewicz CJ, Blair RW, Moodie CA, Isik AI, Erramuzpe A, Kent JD, Goncalves M, DuPre E, Snyder M, Oya H, Ghosh SS, Wright J, Durnez J, Poldrack RA, Gorgolewski KJ. fMRIPrep: a robust preprocessing pipeline for functional MRI. Nat Methods. 2019 Jan;16(1):111-116.

IMPORTANT: (October 2021) The memory and cpu limit flags in fmriprep do not work as intended.
The flags that are used to select an upper bound for fmriprep memory processes and CPUs do not work as intended. Therefore, if an fmriprep job exceeds its memory allocation, the job will hang and D processes will be generated in the compute node where the job is running. We are working on a solution to this but in the meantime we recommend carefully profiling memory consumption for a given type of dataset by running small-scale jobs, then gradually increasing the number of jobs when specific memory requirements have been established. Regarding the CPU flag not working, please use the --exclusive flag if using sbatch or -t auto if using swarm. Please get in touch with HPC staff if you need help with profiling memory requirements or allocating CPUs.

• Fmriprep Documentation
• Github page

• Module Name: fmriprep (see the modules page for more information)
• fmriprep is a scientific pipeline with the potential to overload Biowulf's central filesystem. To avoid filesystem issues we recommend the following:
  1. Limit I/O with central Biowulf directories (e.g., /data, /home) by making use of local disk (/lscratch). You can make use of lscratch to store temporary working directories by directly assigning the temporary work directory of fmriprep (-w option) to /lscratch/$SLURM_JOB_ID (remember to allocate enough space in /lscratch).
  2. Allocate all CPUs in a node. The flag that limits the number of CPUs in MRIQC do not work so please use the --exclusive flag if using sbatch or -t auto if using swarm.
  3. Estimate memory usage. Determine how much memory your jobs will need by benchmarking (see point 5 below). Allocate some extra memory for your jobs as described in the resources described in point 5 below.
  4. Opt out of uploading fmriprep metrics to fmriprep website. You can disable the submission with the --notrack flag.
  5. Profile/benchmark fmriprep jobs: We recommend making sure a given fmriprep job can scale before launching a large number of jobs. You can do this by profiling/benchmarking small jobs (e.g., a swarm of 3 fmriprep commands), then monitoring the jobs by using either the user dashboard or the commands jobhist, sjobs, squeue, and jobload (see biowulf utilities). There are resources prepared by the HPC staff that go over how to monitor your jobs for benchmarking and profiling (video, and slides). Once you have profiled a swarm with a few jobs and you have refined the memory, and walltime requirements, it is probably safe to (gradually) increase the number of fmriprep jobs. For many analyses pipelines one has no way of knowing in advance how much memory will be actually required in an HPC environment. This is why it is very important to profile/benchmark.

Allocate an interactive session and run the program.
Sample session (user input in bold):

[user@biowulf]$ sinteractive
salloc.exe: Pending job allocation 46116226
salloc.exe: job 46116226 queued and waiting for resources
salloc.exe: job 46116226 has been allocated resources
salloc.exe: Granted job allocation 46116226
salloc.exe: Waiting for resource configuration
salloc.exe: Nodes cn3144 are ready for job

[user@cn3144 ~]$ module load fmriprep

[user@cn3144 ~]$ fmriprep -h

[user@cn3144 ~]$ exit
salloc.exe: Relinquishing job allocation 46116226
[user@biowulf ~]$

Create a batch input file (e.g. fmriprep.sh). For example (using test data below):

#!/bin/bash
#SBATCH --job-name=fmriprep
#SBATCH --gres=lscratch:20
#SBATCH --exclusive
#SBATCH --mem=32g
#SBATCH --time=72:00:00

module load fmriprep/20.2.1

tar -C /lscratch/${SLURM_JOB_ID} -xf /usr/local/apps/fmriprep/TEST_DATA/ds001.tar.gz

fmriprep /lscratch/${SLURM_JOB_ID}/ds001 /lscratch/${SLURM_JOB_ID}/fmriprep.out.ds001 \
         participant --participant_label sub-01 -w /lscratch/${SLURM_JOB_ID} \
         --notrack --use-aroma

Submit this job using the Slurm sbatch command.

sbatch fmriprep.sh

Create a swarmfile (e.g. fmriprep.swarm). For example:

fmriprep /data/${USER}/BIDS-dataset/ds001/ /data/$USER/BIDS-dataset/fmriprep.out.ds001 \
         participant --participant_label sub-01 -w /lscratch/${SLURM_JOB_ID} \
         --use-aroma --notrack
fmriprep /data/${USER}/BIDS-dataset/ds002/ /data/$USER/BIDS-dataset/fmriprep.out.ds002 \
         participant --participant_label sub-02 -w /lscratch/${SLURM_JOB_ID} \
         --use-aroma --notrack
fmriprep /data/${USER}/BIDS-dataset/ds003/ /data/$USER/BIDS-dataset/fmriprep.out.ds003 \
         participant --participant_label sub-03 -w /lscratch/${SLURM_JOB_ID} \
         --use-aroma --notrack 

Submit this job using the swarm command.

swarm -f fmriprep.swarm [--gres=lscratch:#] [-g #] -t auto --module fmriprep