AI Wokflow Orchestrator

Orthanc Plugin to simulate an AI Workflow Orchestrator (maintaining workitem tickets/jobs for AI pipeline "workitems"). This is an example implementation for demonstration purposes only at this point. It may evolve into a more mature product over time, but right now, it is 100% work in progress.

How does it work? (tl;dr)

1. Create "workitems" (ie job ticket) via one of two ways:
   1. Uploading a study into Orthanc (triggers will automatically create a new workitem)
   2. Manually create your own worklitem
2. Retrieve workitems via /workitems/ for a full list or workitem/xxx for a specific one
3. Update the state of an existing work item from SCHEDULE to IN PROGRESS, COMPLETED or CANCELED.

---

Usage

Starting up

After cloning this repository, run the following command to start the Orthanc docker with this plugin running:

sudo docker-compose up --build

Afterwards, upload one or more studies into Orthanc. After 5 seconds (the configured waiting period for a "stable" study), you can perform an HTTP GET to:

http://localhost:8042/ai-orchestrator/workitems

Which should return a list of workitems (same number as the number of studies you uploaded). The workitems are DICOM JSON objects.

Workitem Manipulation

Get an individual workitem

Perform an HTTP GET to

http://localhost:8042/ai-orchestrator/workitems/xxx

Replace xxx with the workitem's ID that you got from the listing.

Changing a workitem's state

Perform an HTTP PUT to

http://localhost:8042/ai-orchestrator/workitems/xxx/state

Replace xxx with the workitem's ID that you got from the listing. The body of your request should be identical to work item you are trying adjust, except for the 00741000 attribute (aka Procedure Step State). Which should have one of the following values:

• SCHEDULED
• IN PROGRESS
• CANCELED
• COMPLETED

Manually creating a workitem

Perform an HTTP POST to

http://localhost:8042/ai-orchestrator/workitems/

The body of your request should be a DICOM JSON object, similar to what you would get when doing an HTTP GET to /workitems/xxx (i.e. you can take that as a template and modify it to fit your needs).

---

Credits

• Brad Genereaux - The mastermind behind the idea and developer of the first, Java-based, proof-of-concept
• Mohannad Hussain - Developed this iteration of the AI Orchestrator as an Orthanc Python Plugin
• Society for Imaging Informatics in Medicine (SIIM) - For fostering innovation in Imaging IT/Informatics and sponsoring this project

---

Contribution

• Code: Fork this repository, make your changes, then submit a poll request.
• Other: Contact Mohannad Hussain