12.23. Clara Deploy DICOM Series Selection Pipeline
This asset requires the Clara Deploy SDK. Follow the instructions on the Clara Ansible page to install the Clara Deploy SDK.
This pipeline is one of the reference pipelines provided in the Clara Deploy SDK, provided to demonstrate how DICOM studies can be parsed with metadata preserved in standards-based Clara Deploy JSON format, the relevant series matched with simple series matching rules, and the converted image file(s) selected for the matched series. This pipeline serves as the inspiration or basis to develop pipelines which can intelligently validate input DICOM studies and decide the correct AI algorithms for carrying out data processing.
The pipeline takes as input a set of DICOM instances, from the same or different studies. The first operator in the pipeline, DICOM Parser, performs the following operations,
- parses all the instances and organizes them into DICOM object hierarchy, i.e. study/series/instances. Key DICOM metadata are extracted and saved in standards-based Clara Deploy JSON format, in file
- coverts appropriate DICOM series, e.g. CT and MR image series, to volume images in MetaImage format or NIfTI format per runtime configuration. Series instance UID to image file mapping is saved in the file
series-images.json, along with the image files.
The next operator in the pipeline, Series Selector, continue the processing with the following operations,
- loads the simple series matching rules, and tries to match the series with the DICOM metadata;
- For matched series, finds the mapped image file in the
series-images.json, and saves the matched image file paths in
It is envisioned that other operators, e.g. AI inference operator, can then be added to load the selected image file(s) according to
selected-images.json with the certainty that the correct image is used.
This pipeline is defined in the Clara Deploy pipeline definition language, and for brevity, only the DICOM Parser and Series Selector operators are included.
The following are the details of the definition, with comments describing each operator’s functions as well as IO, while the actual Docker image name and its tag will be updated when published.
api-version: 0.4.0 name: series-selection-pipeline operators: # DICOM Parser operator # Input: '/input' mapped directly to the input of the pipeline, which is populated by the DICOM Adaptor. # Output:'/output' for saving the DICOM metadata in Standards-based Clara Deploy JSON format, # the converted volume image files, and the series instance UID to image file mappings in JSON. - name: dicom-parser description: Converts DICOM instances into image files, one per DICOM series, along with metadata files. container: image: clara/dicom-parser tag: latest input: - path: /input output: - path: /output # series-selection operator # Input: `/input` containing Clara Deploy DICOM metadata, series to image mapping file, and optionally image files. # Output: `/output` containing metadata in JSON specifying selected image file(s) for matched series. # `/publish` containing original and segmented volume images, MHD format, # along with rendering configuration file. - name: series-selection description: Selects the image file(s) for the matched series using simplistic series matching rules. container: image: clara/series-selector tag: latest input: - from: dicom-parser path: /input output: - path: /output name: selections
Please refer to the How to Run a Reference Pipeline section to learn how to register a pipeline, configure the DICOM Adapter, and execute the pipeline.
An End User License Agreement is included with the product. By pulling and using the Clara Deploy asset on NGC, you accept the terms and conditions of these licenses.
Release Notes, the Getting Started Guide, and the SDK itself are available at the NVIDIA Developer forum: (https://developer.nvidia.com/clara).
For answers to any questions you may have about this release, visit the NVIDIA Devtalk forum: (https://devtalk.nvidia.com/default/board/362/clara-sdk/).