hDsProc : Preprocess large-scale sleep fMRI data using AFNI and SLURM on the Biowulf cluster.

AUTHOR: [email protected]

A fairly compact and efficient processing pipeline implemented in Bash and Python using afni_proc.py and the batch processing system SLURM on Biowulf the HPC cluster at the NIH.

USAGE

Examples: see h_15xx_proc.sh

$> hDsAp5_swob_v5a?.sh -h

$> cd output/directory
$> hDsAp5_swob_v5a?.sh -t sleep -S All /input/data/dir

hDsAp5_swob_* will copy the folder structure and links to the sub-*_task-sleep_*.nii files found in /input/data/dir. It will also create run_swarm.sh and swarm_jobs.sh in the output dir.

run_swarm submits the jobs in swarm_jobs.sh to SLURM. Each compute node executes hDsAp5_swob_v5a?.sh -J All on one experiment in one of the output directories. The parameter All selects a subset of parameters for afni_proc.py. See also hDsAp5_swob_v5a?.sh -h.

Output

• Log files: .../output/directory/log/slurm-*.out

• A great number of AFNI output files are found in: ../out/dir/sub-00001/ses-1/run-$ExId/$ExId.results/

  • View quality-control output in ../$ExId.results/QC*/index.html
  • Epi_Mc.nii Preprocessed fMRI time series, aligned to Talairach space by default.
  • McPar.1D (= dfile*.1D) : time*6 motion parameters
  • AnaT1_1mm.nii : T1W anatomy, aligned to Talairach by non-lin. warp