Skip to content

imgnet CLI#

imgnet#

IMGNET CLI - toolkit for facilitating access to standardized medical imaging datasets for cancer research and clinical AI applications.

Usage:

imgnet [OPTIONS] COMMAND [ARGS]...

Options:

Name Type Description Default
--quiet, -q boolean Suppress all logging except errors, overrides verbosity options. False
--verbose, -v integer range (0 and above) Increase verbosity of logging, overrides environment variable. (0-3: ERROR, WARNING, INFO, DEBUG). 0
--version boolean Show the version and exit. False
-h, --help boolean Show this message and exit. False

Subcommands

  • collections: Explore datasets available in the ImgNet index.
  • download: Download medical imaging data from an ImgNet query manifest.
  • query: Query medical imaging datasets and save matching DICOM series metadata.
  • update-index: Download and update the local ImgNet dataset index.

collections#

Explore datasets available in the ImgNet index.

Collections are indexed medical imaging datasets that can be searched, inspected, and downloaded using ImgNet commands.

Usage:

imgnet collections [OPTIONS] COMMAND [ARGS]...

Options:

Name Type Description Default
--help, -h boolean Show this message and exit. False

Subcommands

  • info: Display detailed information about an indexed collection.
  • overview: Display a visual overview of the ImgNet dataset index.
  • summary: Display a summary of all indexed collections.

info#

Display detailed information about an indexed collection.

Shows metadata such as the data source, file format, image count, modalities, body regions, and collection description.

Use --tags to display the metadata fields available for filtering when building ImgNet queries.

Usage:

imgnet collections info [OPTIONS] COLLECTION_NAME

Options:

Name Type Description Default
--help, -h boolean Show this message and exit. False
--tags, -t boolean Show metadata fields available for querying, grouped by modality. False
--json boolean Output collection metadata as JSON instead of a formatted table. False

overview#

Display a visual overview of the ImgNet dataset index.

Shows aggregate statistics across all collections, including: number of collections, total images, storage size, data sources, file formats, modality distributions, and largest collections.

Usage:

imgnet collections overview [OPTIONS]

Options:

Name Type Description Default
--help, -h boolean Show this message and exit. False
--update, -u boolean Rebuild the collections summary before generating the overview. False

summary#

Display a summary of all indexed collections.

Shows basic information including collection size, image count, modalities, file formats, and data sources.

Use --update to rebuild the summary from the collection metadata.

Usage:

imgnet collections summary [OPTIONS]

Options:

Name Type Description Default
--help, -h boolean Show this message and exit. False
--update, -u boolean Rebuild the collections summary before generating the overview. False

download#

Download medical imaging data from an ImgNet query manifest.

The manifest should be generated using imgnet query and contains a list of imaging series to download from indexed datasets.

By default, downloaded files are stored in the srcdata directory inside the output directory. Use --process to convert downloaded DICOM files into NIfTI format using med-imagetools. Processed files are stored in the procdata directory.

MANIFEST: Path to a CSV manifest generated by imgnet query.

Usage:

imgnet download [OPTIONS] MANIFEST

Options:

Name Type Description Default
--help, -h boolean Show this message and exit. False
--output-dir, -o directory Directory where downloaded data will be stored. If not provided, a timestamped directory will be created in the current working directory. None
--process, -p boolean Convert downloaded DICOM files to NIfTI format using med-imagetools. False
--delete-srcdata boolean Delete downloaded DICOM source files after successful processing. This frees disk space but cannot be undone. False

query#

Query medical imaging datasets and save matching DICOM series metadata.

Searches across indexed imaging collections to find series that match your criteria. Results are saved as a CSV file that can be piped into imgnet download to retrieve the actual image files.

Output files in : - query_results.csv : List of matching series with their metadata - valid_query.json : The query used (can be reused with --input-path) - valid_query_schema.json: JSON schema for the query format

QUERY LOGIC: All specified filters are combined with AND logic. For example, using --collection 4D-Lung --modality CT --rules "CT(PatientAge>50)" will only match CT series from the 4D-Lung collection where patient age > 50.

Rules provide the most powerful filtering and support: - Operators: ==, !=, <=, >=, <, > - Logical: AND, OR, and parentheses for grouping - Values: strings (quoted), numbers, or unquoted words

EXAMPLES: # Query all CT scans from the 4D-Lung collection imgnet query -c 4D-Lung -m CT

# Query multiple collections for MR and CT, save to custom directory imgnet query -o ./my_query -c 4D-Lung -c Adrenal-ACC-Ki67-Seg -m MR -m CT

# Find female patients over 50 with CT chest scans imgnet query -r 'CT(PatientSex=="F" AND PatientAge>50 AND StudyDescription=="CHEST")'

# Find either CT or MR studies for patients under 30 imgnet query -r 'CT(PatientAge<30) OR MR(PatientAge<30)'

# Re-run a previous query imgnet query -i ./previous_query/valid_query.json

Usage:

imgnet query [OPTIONS]

Options:

Name Type Description Default
--help, -h boolean Show this message and exit. False
--output-dir, -o directory Directory to save query results (default: ./query_results). None
--input-path, -i file Re-run a previous query by loading a saved valid_query.json file. None
--collection, -c text Include only files from this dataset collection. Can be specified multiple times. Use 'all' to query every collection. None
--modality, -m text Include only files of this imaging modality (e.g., CT, MR, RTSTRUCT, SEG). Can be specified multiple times. None
--file-type, -ft choice (dicom | nifti) Include only files in this format: 'dicom' or 'nifti' (default: both). None
--rules, -r text Filter series by metadata using boolean expressions. Format: modality(condition AND/OR condition). Example: -r 'CT(PatientAge>50 AND StudyDescription!="CHEST")' Example: -r 'MR(SeriesDescription=="T2_AXIAL")' Example: -r 'CT(PatientSex=="F" OR PatientAge<30)' None

update-index#

Download and update the local ImgNet dataset index.

The index contains metadata describing available imaging collections, including information required for searching and downloading datasets.

By default, the index is stored in the ImgNet user data directory. Use --path to store or update the index in a custom location.

Usage:

imgnet update-index [OPTIONS]

Options:

Name Type Description Default
--path, -p path Path where the ImgNet indexed datasets will be stored. Defaults to $IMGNET_INDEX_DIR/indexed_datasets or the platform user data directory if the environment variable is not set. None
--help boolean Show this message and exit. False