voxelops 0.3.2

Clean, simple neuroimaging pipeline automation for brain banks

Clean, simple neuroimaging pipeline automation for brain banks.

Brain banks need to process neuroimaging data consistently, reproducibly, and auditably. VoxelOps makes that simple by wrapping Docker-based neuroimaging tools into clean Python functions that return plain dicts – ready for your database, your logs, and your peace of mind.

Overview

docs
tests, CI & coverage
version
styling
license

Features

• Simple Functions – No classes, no inheritance – just run_*() functions that return dicts

• Clear Schemas – Typed dataclass inputs, outputs, and defaults for every procedure

• Reproducibility – The exact Docker command is stored in every execution record

• Database-Ready – Results are plain dicts, trivial to save to PostgreSQL, MongoDB, or JSON

• Brain Bank Defaults – Define your standard parameters once, reuse across all participants

• Comprehensive Logging – Every run logged to JSON with timestamps, duration, and exit codes

• Validation Framework – Pre- and post-execution validation with detailed reports

• Audit Trail – Full audit logging for every procedure run

Installation

pip install voxelops

For development:

git clone https://github.com/yalab-devops/VoxelOps.git
cd VoxelOps
pip install -e ".[dev]"

Requirements: Python >= 3.10, Docker installed and accessible.

Quick Start

Basic (direct execution):

from voxelops import run_qsiprep, QSIPrepInputs

inputs = QSIPrepInputs(
    bids_dir="/data/bids",
    participant="01",
)

result = run_qsiprep(inputs, nprocs=16)

print(f"Completed in: {result['duration_human']}")
print(f"Outputs: {result['expected_outputs'].qsiprep_dir}")
print(f"Command: {' '.join(result['command'])}")

With validation and audit logging (recommended):

from voxelops import run_procedure, QSIPrepInputs

inputs = QSIPrepInputs(
    bids_dir="/data/bids",
    participant="01",
)

result = run_procedure("qsiprep", inputs)

if result.success:
    print(f"Completed in {result.duration_seconds:.1f}s")
else:
    print(f"Failed: {result.get_failure_reason()}")

# Save complete audit trail to your database
db.save_procedure_result(result.to_dict())

Available Procedures

Procedure	Purpose	Function	Execution
HeudiConv	DICOM to BIDS conversion	run_heudiconv()	Docker
QSIPrep	Diffusion MRI preprocessing	run_qsiprep()	Docker
QSIRecon	Diffusion reconstruction & connectivity	run_qsirecon()	Docker
QSIParc	Parcellation via parcellate	run_qsiparc()	Python (direct)

Brain Bank Standards

Define your standard parameters once, use them everywhere:

from voxelops import run_qsiprep, QSIPrepInputs, QSIPrepDefaults

BRAIN_BANK_QSIPREP = QSIPrepDefaults(
    nprocs=16,
    mem_mb=32000,
    output_resolution=1.6,
    anatomical_template=["MNI152NLin2009cAsym"],
    docker_image="pennlinc/qsiprep:latest",
)

for participant in participants:
    inputs = QSIPrepInputs(bids_dir=bids_root, participant=participant)
    result = run_qsiprep(inputs, config=BRAIN_BANK_QSIPREP)
    db.save_processing_record(result)

Validation & Audit

run_procedure() wraps any runner with pre-validation, post-validation, and a full audit trail:

from voxelops import run_procedure, HeudiconvInputs, HeudiconvDefaults

inputs = HeudiconvInputs(
    dicom_dir="/data/dicoms",
    participant="01",
    session="baseline",
)
config = HeudiconvDefaults(heuristic="/code/heuristic.py")

result = run_procedure("heudiconv", inputs, config)

# result.pre_validation  -- ValidationReport before execution
# result.post_validation -- ValidationReport after execution
# result.audit_log_file  -- path to the JSON audit log

Logging

All runners accept an optional log_dir parameter. When provided, an execution JSON log is written alongside any audit logs. The log directory defaults to <output_dir>/../logs derived from the inputs.

result = run_qsiprep(inputs, log_dir="/data/logs/qsiprep")

Documentation

Full documentation is available at voxelops.readthedocs.io.

License

MIT License – see the LICENSE file for details.