BRAINSPydra

Using BRAINSPydra

Using BRAINSPydra to run AutoWorkup using the Pydra dataflow engine requires the following steps:

1. SSH into the SGE cluster (at the University of Iowa, this is Argon)
2. Activate a python environment with python version 3.7+
3. Clone this repository and install requirements pip install -r REQUIREMENTS.txt
4. Define config_environment to set parameters for individual steps in AutoWorkup and config_experimental to set parameters for the computing environment (examples: config_environment_argon.json and config_experimental_argon.json)
5. Generate an array job of AutoWorkup pipelines to avoid the Pydra overhead of submitting many sessions through the same pipeline script (session_count recommended to be 10). Further instruction is given in Generating Pipeline Jobs.
6. ssh into Argon
7. Submit the array job of pipelines to the Sun Grid Engine. Further instruction is given in Running the Pipeline.

Generating Pipeline Jobs

Parameters

positional arguments:
  bids_path             The path to the top level of the bids dataset
  best_image_table      Path to the tsv file containing information on the best series for each
                        session
  data_dictionary_dir   Path the the directory to store the generated input_data_dictionary__.json
                        files
  pipeline_script       Path to the python pipeline script to run BAW using pydra
  experimental_config   Path to the experimental configuration json file defining settings for each of
                        the BRAINSTools applications
  environmental_config  Path to the environmental configuration json file defining overall settings
                        for the system being used

optional arguments:
  -h, --help            show this help message and exit
  --output_job_path OUTPUT_JOB_PATH
                        Output path to the job file that will be run to execute the pipeline on all
                        the input data dictionary files
  --max_sessions_per_file MAX_SESSIONS_PER_FILE
                        The number of sessions to be recorded per json file (-1 for all sessions in
                        one file)
  --session_count SESSION_COUNT
                        The number of sessions from which to have data extracted
  --original_sessions_list_file ORIGINAL_SESSIONS_LIST_FILE
                        The path to a list file where each line is a session (sess-#####) processed in
                        the original BAW predicthd run
  --bad_sessions_list BAD_SESSIONS_LIST
                        The path to a list file where each line is a session (sess-####) that should
                        not be processed

Example

python3.8 extract_input_data.py \
/Shared/johnsonhj/ReferenceData/PREDICTHD_BIDS_DEFACE \
/Shared/johnsonhj/ReferenceData/PREDICTHD_BIDS_DEFACE/phenotype/bids_best_image_table.tsv \
$(pwd)/input_data_dictionaries \
$(pwd)/preliminary_pipeline6.py \
$(pwd)/config_experimental_argon.json \
$(pwd)/config_environment_argon.json \
--output_job_path pipeline_20.job \
--max_sessions_per_file 10 \
--original_sessions_list_file ./my_sessions.list \
--session_count 20

Running the Pipeline

Parameters

usage: preliminary_pipeline6.py [-h] config_experimental config_environment input_data_dictionary

Move echo numbers in fmap BIDS data to JSON sidecars

positional arguments:
  config_experimental   Path to the json file for configuring task parameters
  config_environment    Path to the json file for setting environment config parameters
  input_data_dictionary
                        Path to the json file for input data

optional arguments:
  -h, --help            show this help message and exit

Example

qsub -pe smp 1 -q HJ -t 1-2 pipeline_20.job

Other Notes

Rerunning the Pipeline

If all array job entries in pipeline_20.job complete, but not all sessions are fully processed, rerun the pipeline. Before rerunning, run the command rm *.lock in the cache_dir given in the config_experimental JSON file. Then, run the same command used to run the pipeline, as in Running the Pipeline/Example