I just discovered OHIF, and I’m trying to understand how custom extensions work.
We’d like to add a ML application to the viewer (for example, performing automatic segmentation of organs on the image), but we only code in Python.
Does anyone know if it is possible to connect a Docker container to the app and set up an API to manage requests from/to the web app?
Is there any example available?
Thank you very much for your support - and for making OHIF opensource!
I would suggest that you look into XNAT’s branch of the OHIF Viewer, and implement the integrated NVIDIA Clara Train SDK tooling which they’ve created to work with OHIF. This is a tool which uses machine learning and AI to create segmentations like you are looking for.
More info on the tooling here
and setup is here
the version of the OHIF viewer with these integrated tools can be downloaded here
You can also take a look at ProstateCancer.ai project which performs uses AI classifier for malignancy prediction. The source code (which is based on OHIF-v1, but you will get some idea) is here
Hello Shane and alireza, thank you very much for your suggestions, I’ll surely have a look into those two interesting projects!
Unfortunately XNAT ML , doesn’t have any method to deploy the AI solution. Only prostate.ai have one but that can’t be universalized either as it’s based on prostate based template.
We are also looking at ways to deploy AI based solution through OHIF and currently running XNAT training data. Would your team be interested in collaboration?
Unfortunately XNAT ML , doesn’t have any method to deploy the AI solution.
It looks like I pasted the wrong link for the setup information in my post above, my mistake.
Here is the solution provided by XNAT to deploy their Dockerized Clara Train SDK: https://wiki.xnat.org/documentation/xnat-ohif-viewer/xnat-ohif-plugin-installation-and-administration-124813344.html
I deployed the setup on AWS using the hardware suggested by NVIDIA, which can be seen here Clara Train - Getting started with a Cloud Service Provider — Clara Train SDK v4.0 documentation
Hope this helps. If you have any questions about the setup, a great place to ask is the XNAT google group found here https://groups.google.com/g/xnat_discussion
Just revisiting this issue again.
We tried to use prostatecancer ai thing. But hitting road block. Is there a simplier solution out there to implement the ohif viewer based deployment?