CLI Reference#

The LIOM Toolkit ships seven liom-* command-line tools, registered as console scripts in pyproject.toml under [project.scripts]. Each tool is a thin wrapper over a library function: the script module defines _build_argument_parser() (which builds the argparse tree without importing heavy dependencies) and main() (which lazy-imports the heavy deps and wires an optional --dask_scheduler argument to dask_client_manager.set_client(...) before calling domain logic).

The reference below is auto-generated by sphinx-argparse from each script’s live _build_argument_parser() function, so the documented options always match the code — there is no hand-maintained argument table to drift.

Run any tool with --help for the authoritative options list:

liom-convert-hdf5-to-zarr --help

Conversion#

liom-convert-hdf5-to-zarr#

CLI: convert an HDF5 volume to an OME-Zarr store.

usage: liom-convert-hdf5-to-zarr [-h] [--dask_scheduler DASK_SCHEDULER]
                                 [--n_workers N_WORKERS]
                                 [--log-level {DEBUG,INFO,WARNING,ERROR}]
                                 [--resume] [--scales SCALES SCALES SCALES]
                                 [--chunks CHUNKS CHUNKS CHUNKS]
                                 input_file output_file

Positional Arguments#

input_file: Full path to the input HDF5 file
output_file: Full path to the output Zarr file

Named Arguments#

--dask_scheduler

Network address of the dask scheduler (default: local cluster)

--n_workers

Number of Dask local-cluster workers (default: min(cpu_count()-1, 8))

--log-level

Possible choices: DEBUG, INFO, WARNING, ERROR

Logging level (default: INFO)

Default: 'INFO'

--resume

Resume from checkpoint if available (default: False)

Default: False

--scales

Scales (voxel size) for the Zarr dataset (default=(6.5, 6.5, 6.5))

Default: (6.5, 6.5, 6.5)

--chunks

Chunk size for the Zarr dataset (default=(128, 128, 128))

Default: (128, 128, 128)

Mask generation#

liom-create-mask#

CLI: create a brain mask for an OME-Zarr volume and write it to disk.

usage: liom-create-mask [-h] [--dask_scheduler DASK_SCHEDULER]
                        [--n_workers N_WORKERS]
                        [--log-level {DEBUG,INFO,WARNING,ERROR}] [--resume]
                        [--scales SCALES SCALES SCALES]
                        [--chunks CHUNKS CHUNKS CHUNKS] [--fill_holes]
                        input_file

Positional Arguments#

input_file: Full path to the input zarr file

Named Arguments#

--dask_scheduler

Network address of the dask scheduler (default: local cluster)

--n_workers

Number of Dask local-cluster workers (default: min(cpu_count()-1, 8))

--log-level

Possible choices: DEBUG, INFO, WARNING, ERROR

Logging level (default: INFO)

Default: 'INFO'

--resume

Resume from checkpoint if available (default: False)

Default: False

--scales

Scales (voxel size) for the Zarr dataset (default=(6.5, 6.5, 6.5))

Default: (6.5, 6.5, 6.5)

--chunks

Chunk size for the Zarr dataset (default=(128, 128, 128))

Default: (128, 128, 128)

--fill_holes

Fill holes in the mask (default: False)

Default: False

Segmentation#

liom-segment-2d#

Registration#

liom-align-annotations#

CLI: align annotations to a target volume.

usage: liom-align-annotations [-h] [--dask_scheduler DASK_SCHEDULER]
                              [--n_workers N_WORKERS]
                              [--log-level {DEBUG,INFO,WARNING,ERROR}]
                              [--resume] [--resolution RESOLUTION]
                              [--rigid_type RIGID_TYPE]
                              [--deformable_type DEFORMABLE_TYPE]
                              target_volume mask template atlas data_dir

Positional Arguments#

target_volume: Path to the target volume image (NIfTI/NRRD)
mask: Path to the mask image (NIfTI/NRRD)
template: Path to the template image (NIfTI/NRRD)
atlas: Path to the atlas/annotation image (NIfTI/NRRD)
data_dir: Directory for intermediary files

Named Arguments#

--dask_scheduler

Network address of the dask scheduler (default: local cluster)

--n_workers

Number of Dask local-cluster workers (default: min(cpu_count()-1, 8))

--log-level

Possible choices: DEBUG, INFO, WARNING, ERROR

Logging level (default: INFO)

Default: 'INFO'

--resume

Resume from checkpoint if available (default: False)

Default: False

--resolution

Atlas resolution in micron, must be 10/25/50/100 (default=25)

Default: 25

--rigid_type

Rigid registration type (default=’Similarity’)

Default: 'Similarity'

--deformable_type

Deformable registration type (default=’SyN’)

Default: 'SyN'

liom-build-template#

CLI: build a groupwise registration template for a set of OME-Zarr brains.

usage: liom-build-template [-h] [--dask_scheduler DASK_SCHEDULER]
                           [--n_workers N_WORKERS]
                           [--log-level {DEBUG,INFO,WARNING,ERROR}] [--resume]
                           [--resolution_level RESOLUTION_LEVEL]
                           [--template_resolution TEMPLATE_RESOLUTION]
                           [--atlas_resolution ATLAS_RESOLUTION]
                           [--iterations ITERATIONS]
                           output_file zarr_files [zarr_files ...]
                           brain_names [brain_names ...]

Positional Arguments#

output_file: Path where the output template is saved (NIfTI/NRRD)
zarr_files: Paths to the input OME-Zarr brain files
brain_names: Brain names for saving pre-registered images

Named Arguments#

--dask_scheduler

Network address of the dask scheduler (default: local cluster)

--n_workers

Number of Dask local-cluster workers (default: min(cpu_count()-1, 8))

--log-level

Possible choices: DEBUG, INFO, WARNING, ERROR

Logging level (default: INFO)

Default: 'INFO'

--resume

Resume from checkpoint if available (default: False)

Default: False

--resolution_level

OME-Zarr resolution level to load (default=3)

Default: 3

--template_resolution

Template resolution in micron (default=50)

Default: 50

--atlas_resolution

Allen atlas resolution in micron (default: template_resolution)

--iterations

Number of groupwise registration iterations (default=15)

Default: 15

Statistics#

liom-compute-slice-metrics#

CLI: compute per-region vessel morphometric statistics for a brain slice.

usage: liom-compute-slice-metrics [-h] [--dask_scheduler DASK_SCHEDULER]
                                  [--n_workers N_WORKERS]
                                  [--log-level {DEBUG,INFO,WARNING,ERROR}]
                                  [--resume] [--voxel_size VOXEL_SIZE]
                                  output_dir image mask_file vessel_mask_file
                                  region_map_file vessel_exclude_file

Positional Arguments#

output_dir: Directory to save the output metrics to
image: Label (filename/identifier) of the brain slice, used in the output DataFrame
mask_file: Path to the tissue mask image (TIFF/PNG)
vessel_mask_file: Path to the vessel mask image (TIFF/PNG)
region_map_file: Path to the region map image (TIFF/PNG)
vessel_exclude_file: Path to the vessel-exclude mask image (TIFF/PNG)

Named Arguments#

--dask_scheduler

Network address of the dask scheduler (default: local cluster)

--n_workers

Number of Dask local-cluster workers (default: min(cpu_count()-1, 8))

--log-level

Possible choices: DEBUG, INFO, WARNING, ERROR

Logging level (default: INFO)

Default: 'INFO'

--resume

Resume from checkpoint if available (default: False)

Default: False

--voxel_size

Voxel size in micron (default=0.65)

Default: 0.65

Deep learning#

liom-train-model#

CLI: train the vessel segmentation U-Net on an OME-Zarr dataset.

usage: liom-train-model [-h] [--dask_scheduler DASK_SCHEDULER]
                        [--n_workers N_WORKERS]
                        [--log-level {DEBUG,INFO,WARNING,ERROR}] [--resume]
                        [--output_train OUTPUT_TRAIN] [--epochs EPOCHS]
                        [--batch_size BATCH_SIZE]
                        [--learning_rate LEARNING_RATE]
                        [--wandb_entity WANDB_ENTITY]
                        [--wandb_project WANDB_PROJECT]
                        [--pretrained_artifact PRETRAINED_ARTIFACT]
                        [--wandb_mode WANDB_MODE]
                        dataset_file node_name

Positional Arguments#

dataset_file: Path to the OME-Zarr dataset file
node_name: Name of the node in the zarr file

Named Arguments#

--dask_scheduler

Network address of the dask scheduler (default: local cluster)

--n_workers

Number of Dask local-cluster workers (default: min(cpu_count()-1, 8))

--log-level

Possible choices: DEBUG, INFO, WARNING, ERROR

Logging level (default: INFO)

Default: 'INFO'

--resume

Resume from checkpoint if available (default: False)

Default: False

--output_train

Output directory for training artifacts (default=’training’)

Default: 'training'

--epochs

Number of training epochs (default=62)

Default: 62

--batch_size

Training batch size (default=35)

Default: 35

--learning_rate

Optimizer learning rate (default=0.003673)

Default: 0.003673

--wandb_entity

wandb entity/team (default: wandb user default)

--wandb_project

wandb project name (default: wandb default project)

--pretrained_artifact

wandb artifact path for pretrained weights (default: train from scratch)

--wandb_mode

wandb mode: online/offline/disabled (default=’offline’)

Default: 'offline'

Note

The liom-train-model and liom-segment-2d (U-Net mode) tools require the ai extra (pip install liom-toolkit[ai]). The liom-build-template and liom-align-annotations tools require the antspy extra (pip install liom-toolkit[antspy]). See Getting Started for the full extras matrix.