Components of the XNAT ML-Beta Release

The XNAT ML (Beta) release is a highly customized and prepackaged XNAT platform solution that "ships" with plugins and extra components preinstalled, using either Docker-Compose or Kubernetes to bundle these components together. If you are skilled in DevOps and want to roll out your own version of this release on the technology stack of your choice, this guide will help you identify and install the relevant components.

Supported Distributions:

Components of This Release (And Where To Find Them)

ComponentRelease VersionDescriptionSource
XNAT war file - "feature/ML" buildML-BETA

The core XNAT war file is a built on a pre-released version of XNAT 1.8, which has several notable differences from XNAT 1.7. Migrating legacy XNAT data to this release may have challenges.

  • Java 8
  • PostgreSQL 12
  • Support for Event Service, which is based on Spring v4 and Reactor v2
Download: XNAT Features/ML war file
"Clara" PluginML-BETAThis plugin contains the core Machine Learning dashboard and datatypes, and was originally built to support the NVIDIA Clara set of medical imaging AI models.

Download: xnatx-clara-plugin

Data Collection PluginML-BETAThis plugin contains support for the generic concept of "data collections", which can take many forms. It also includes specific support for "tagged resource maps", which are used to transform a list of image sessions into a list of resource files for model training.Download: xnatx-collection-plugin
Container Service PluginML-BETAThis version of the Container Service plugin has been modified to work with Event Services, and has a few other modifications for machine learningDownload: container-service
Batch Launch PluginML-BETAThis plugin, developed by Radiologics, provides a dashboard for launching and managing processing workflowsDownload: batch-launch-plugin
XNAT-OHIF Viewer PluginML-BETAThis version of the XNAT-OHIF Viewer is a pre-release version 3, which adopts a modernized backend and improves the association of annotations with individual scans as XNAT objectsDownload: xnat-ohif-viewer
DICOM Query-Retrieve Plugin ML-BETAThe DQR plugin is slated for release concurrent with XNAT 1.8 and this version of it is feature-complete. It allows XNAT to connect to a PACS and query it for sessions, then pull sessions over.Download: dqr-plugin
ActiveMQ5.14.3ActiveMQ is a JMS messaging queuing service that is normally bundled into an XNAT core distribution, but is run as a separate container here.Download: ActiveMQ
Traefik Proxy Server:latest(Optional) The Traefik edge-routing proxy server is used to handle URL routing to containers running TensorBoard, which allows XNAT to peek into running model training. This could be replaced by some other URL routing scheme as needed.Download: Traefik
Orthanc PACS1.4.2(Optional) An Orthanc PACS has been distributed with this beta package to demonstrate connectivity with the DQR plugin. Orthanc runs in a separate Docker image, and is preconfigured to be aware of your XNAT. You can access the web front-end of your PACS system at http://localhost/pacs.Download: Orthanc

Additional Components

These components are not shipped with the XNAT ML (Beta) release, but they are required for completing elements of the demonstration workflow.

Docker Container Images:

  • xnat/clara-train-v2:0.6
  • xnat/clara-train-v3:dev
  • xnat/demo-tensorboard:latest
  • xnat/rtstruct2composite_mask:v2

Optional Container Images:

  • xnat/dcm2niix:latest

See Install Docker Images and Commands

$label.name