Command-Line Interface

OncoPrep provides four CLI commands:

Command	Purpose
oncoprep	Main preprocessing + segmentation pipeline
oncoprep-convert	DICOM → BIDS conversion
oncoprep-docker	Run OncoPrep inside a Docker container
oncoprep-models	Manage segmentation model images

oncoprep

The main pipeline command. Follows the BIDS-Apps convention:

oncoprep <bids_dir> <output_dir> <analysis_level> [options]

Positional arguments

Argument	Description
bids_dir	Root folder of a valid BIDS dataset
output_dir	Output path for derivatives and reports
analysis_level	participant (subject-level) or group (cohort-level ComBat)

Key options

BIDS filtering

--participant-label LABEL [LABEL ...]
    Subject identifiers (sub- prefix optional)

--session-label LABEL [LABEL ...]
    Session identifiers (ses- prefix optional)

--bids-filter-file PATH
    JSON file with custom pybids query filters

Performance

--nprocs N          Number of CPUs (default: all available)
--omp-nthreads N    Threads per process (default: auto)
--mem-gb N          Memory limit in GB
--low-mem           Trade disk for memory

Workflow configuration

--output-spaces SPACE [SPACE ...]
    Template spaces (default: MNI152NLin2009cAsym)

--skull-strip-backend {ants,hdbet,synthstrip}
    Skull-stripping method (default: ants)

--registration-backend {ants,greedy}
    Registration method (default: ants)

--deface
    Remove facial features for privacy

Segmentation

--run-segmentation    Enable tumor segmentation
--default-seg         Use single default model (CPU-friendly)
--seg-model-path PATH Custom model path
--no-gpu              Force CPU-only execution
--container-runtime {auto,docker,singularity,apptainer}
    Container runtime (default: auto)
--seg-cache-dir PATH  Pre-downloaded model cache directory

Radiomics

--run-radiomics       Enable feature extraction (implies --run-segmentation)

ComBat Harmonization (group-level)

These flags are used with analysis_level = group. See Group-Level ComBat Harmonization for a full guide.

--combat-batch CSV    CSV mapping subjects to scanner/site batches
                      (columns: subject_id, batch, optional covariates)

--combat-parametric   Use parametric empirical Bayes priors (default)

--combat-nonparametric
                      Use non-parametric empirical Bayes

--generate-combat-batch
                      Auto-generate batch CSV from BIDS JSON sidecars
                      (Manufacturer, ManufacturerModelName,
                      MagneticFieldStrength, age, sex)

VASARI

--run-vasari          Enable automated VASARI feature extraction and
                      radiology report generation
                      (requires vasari-auto; implies --run-segmentation)

Quality control (temporarily disabled)

--run-qc              [TEMPORARILY DISABLED] MRIQC quality control
                      integration is non-functional in this release.
                      This flag is accepted but ignored.

Other

--work-dir PATH       Working directory (default: ./work)
--reports-only        Generate reports without running workflows
--write-graph         Export workflow DAG as SVG
--stop-on-first-crash Abort on first error
-v / -vv / -vvv      Increase verbosity (debug: -vvv)

oncoprep-convert

DICOM to BIDS conversion using dcm2niix:

oncoprep-convert <dicom_dir> <bids_dir> [options]

--subject ID        Subject identifier
--session ID        Session identifier
--batch             Process all subject directories in dicom_dir

oncoprep-docker

Wrapper to run OncoPrep inside Docker with automatic volume binding and GPU detection:

oncoprep-docker <bids_dir> <output_dir> participant [oncoprep options]

oncoprep-models

Manage pre-downloaded segmentation model container images:

# List available models
oncoprep-models list

# Download all models
oncoprep-models pull -o /path/to/cache --runtime docker

# Download CPU-only models
oncoprep-models pull -o /path/to/cache --runtime docker --cpu-only

# Check download status
oncoprep-models status -o /path/to/cache