Repository navigation
3D Slicer integration of DL models #3
Description
Activity
- changed the title
[-]3D Slicer integration ([/-][+]3D Slicer integration of DL models[/+]on Nov 17, 2021 Just to confirm, we are able to use the current rest API for running inference by just stacking multiple user inputs into one 4D volume and sending it to the server.
I am currently working on integrating everything into 3D Slicer and need to preprocess the data to match the FCN required input volume dimensions etc.
All the above-shown use cases should be covered by stacking the user inputs into a 4D volume
We are planning on integrating our DL models to be used with 3D Slicer
Dependencies
Functional Requirements
support multiple valve types:
support multiple user input variants (user decides/ system autodetects from available data):
MONAILabel limitations (v0.2)
From my understanding, MONAILabel does currently not allow posting multiple user inputs images (multiple frames + labels).
All the inputs to our FCNs are voxel data with the same dimensions/origin/orientation. We could stack all that data (mid-systolic frame + annulus + APC + ASC + PSC) on the client-side and unstack and assign keys on the server-side and proceed with the pre-transforms as planned.