Scripts and documentation for open datasets
| Dataset | fMRI availability |
|---|---|
| ABCD - The Adolescent Brain Cognitive Development Study | Resting + Task |
| HBN - Child Mind Institute: healthy brain network | Resting + Nat Viewing |
| HCP-D - HCP-Development | Resting + Task |
| PNC - Philadelphia neurodevelopmental cohort | Resting + Task |
| NKI - Enhanced Nathan Kline Institute - Rockland Sample | Resting |
| PING - Pediatric Imaging, Neurocognition, and Genetics | Resting |
| ABIDE - Autism Brain Imaging Data Exchange | Resting |
| CoRR - Consortium for Reliability and Reproducibility | Resting |
| ADHD-200 - International Neuroimaging Datasharing Initiative (INDI), the ADHD-200 Sample | Resting |
| NIHPD - The NIH MRI study of normal brain development | NA |
From the PBIL, tar and upload raw bids dataset onto shared project drive. Each site (ie, /ADHD200/Brown/) is done separately to limit size of tar-archives, transfer speeds, and job-management on ComputeCanada. Use the tar_and_rsync_to_CC.sh script from the PBIL, /shared/uher/FORBOW/OpenDatasets/ADHD200/. This script will tar an entire site's raw BIDS dataset, then rsync up to cedar into a Tar_BIDS folder under the Project space (eg, ~/projects/def-ruher/fmri/ADHD200/Tar_BIDS/Brown.tar)
SSH into cedar.computecanada.ca, [email protected] then use the script below to upack the site BIDS tar file from ~/projects/def-ruher/PROJECT/TarBIDS/ into ~/scratch/fmri/PROJECT/SITE/BIDS/. Specify project and site.
cd ~/projects/def-ruher/fmri/
./scripts/fmri_0_unpack_site_bids_archive.sh -p ADHD200 -s Brown
Another option to execute script: bash script-name-here.sh -arg1 -arg2
Then run every subject individually as one fmriprep-slurm job. First pass, add the '--debug' flag to ensure job-scripts are created without any errors. Then remove the flag to submit to slurm-scheduler. This script by default uses NCPUS=8, MemPerCPU=1024MB, OMP-THREADS=NCPUS, with Slurm-WallTime=12hrs. After jobs are submitted use 'sq' command to check the queue status.
./scripts/fmri_1_slurm_fmriprep_batch.sh -p ADHD200 -s Brown -e [email protected] --debug
Depending on resources used, quality of data, and ComputeCanada workload, a certain amount of fmriprep jobs will not finish within the specified WallTime limit (12hrs). These jobs can be re-started easily, from scratch with original outputs deleted, with another script as described below. Each of these jobs will use NCPUS=12. Use '--debug' on the first try to ensure cancelled jobs are properly found and reported. Once satisfied use the '--submit' flag when ready to run. This script safely removes the incomplete results then calls the fmri_1_slurm_fmriprep_batch.sh above to submit the jobs.
./scripts/fmri_2_restart_cancelled_jobs.sh -p ADHD200 -s Brown -e [email protected] --debug
Once fmriprep has completed for all subjects in a site, use this script to archive the derivatives from user ~/scratch/fmri/PROJECT/ (slower-SSD) back over to group-shared tape-drive ~/projects/def-ruher/fmri/PROJECT/Tar_DERIVS/SITE.tar:
./scripts/fmri_3_archive_site_derivs.sh -p ADHD200 -s Brown
From the PBIL, download the completed site derivatives Tar file.
cd /shared/uher/FORBOW/OpenDatasets/ADHD200/ ./download_tar_derivs.sh -p ADHD200 -s Brown
ADHD200 > HBN > ABIDE > Rockland > Corr