XNAT ML Documentation

The XNAT ML Plugin suite was built around the now-deprecated NVIDIA Clara 2.0 style of models and training. These plugins have been deprecated and are not recommended for production use but are still available as open source resources for those who are interested.

The XNAT ML release is a stand-alone package that combines XNAT 1.8 with several platform and plugin additions to enable a baseline of Machine Learning capability. These capabilities were first demonstrated in collaboration with the NVIDIA Clara suite of medical imaging AI tools, at RSNA 2019. Since then, we have been expanding functionality, deepening integrations, and hardening features to get to the point of this release.

Release Notes

Version 1.0 of the XNAT ML Plugin suite was released alongside the release of XNAT 1.8. What has changed? See XNAT ML 1.0 Release Notes

Upgrading from XNAT ML-Beta? See here for data conversion notes: Migrating XNAT ML data from Beta to Production

Getting Started

• Components of the XNAT ML Release

• Installing the XNAT ML Docker Compose Release

Docker Compose Installation Instructions

• Docker Server Setup

• Docker Compose Installation Instructions

• Rebuilding Your Docker Compose Environment

• DevOps Quick Reference for Containerized XNATs

Project Configuration

• Creating a Project and Assigning Roles

• Using the Machine Learning Project Dashboard

Container Service Configuration

• Configure the Container Service Plugin

• Install Docker Images and Commands

• Configure Installed Commands

PACS Configuration & Data Population

• Connecting XNAT to your Demo Orthanc PACS

Performing Training

Workflows

• Overview of supported workflows

  • Model Training: Pre-coordinated Workflow

  • Model Training: Retrospective Workflow

  • Inference (currently unsupported)

Setup Tasks

• Installing a Model

• Creating a Training Config for your Model

• Configuring a Segmentation Mapping Resource for your Model

• Set up an Automated Workflow to Generate Required Model Resources

Data Import and Collection

• Pushing Data from a PACS to your XNAT using the SCP Receiver

• Drawing Contours with the XNAT OHIF Viewer

• Data Conversion for Model Training

• Defining Parameters for your Dataset

• Validating and Saving your Dataset from Project Data

Model Training

• Launching Initial Training

• Launching Finetune Training

• Viewing Training Progress in TensorBoard

• Viewing Training Logs

• Accessing and Downloading Training Results

• Promoting Training Results as a new "Model"

Other Operations

• Deleting Models, Training, and Datasets