Integration of a deep learning plugin in orthanc

Hello Orthanc team,

I am currently working on a tele-radiology project in which I am going to integrate a deep learning model (U-Net) for the detection of anomalies (image of tuberculosis for example) from images in Orthanc through an API.
I’d like to know how I can integrate the model directly into Orthanc using Python, given that I don’t have any knowledge of C or C++. Using an icon, I can select a few images found in Orthanc and the model will tell me if it’s an anomaly or if I can segment it.
Do you have any resources that might be useful?
Thanks for your help!

2 Likes

I’d recommend hiring Orthanc Team for this custom feature: Orthanc - DICOM Server
I believe orthanc python plugin currently only supports adding custom button to legacy Orthanc Explorer, you may want to try adding custom tool to OHIF viewer:
Add a Tool to the Viewer | OHIF

Here’s what I would have done:

  1. Write a web extension that extracts the study ID from the URL of one of the viewers.
  • http://localhost:8042/osimis-viewer/app/index.html?study={studyID}
  • http://localhost:8042/ohif/viewer?url=../studies/{StudyID}/ohif-dicom-json
  1. Use the Python plugin to extend the REST API so it accepts the extracted study ID.
    When web extension is clicked, it passes the StudyID to a newly created API endpoint for Orthanc and opens it in a new tab: http://localhost:8042/SegmentUnet={studyID}
    More information can be found here:
    (Python plugin for Orthanc — Orthanc Book documentation).
  2. Extract images from instances and pass them through a pretrained model.
  3. Display the results.
3 Likes

Hello,

This is typically the kind of scientific questions for which you could envision at collaboration with my research lab at the UCLouvain university.

Kind Regards,
Sébastien-

2 Likes

Okay, thank you very much.
I’ve had to use OHIF as part of my university studies and now I want to explore the resources you’ve shared with me and follow your guidance.
Thank you once again.