coleb888/cartilage-vulnerability-atlas

Topographic Cartilage Vulnerability Atlas - Shiny App

This Shiny for Python app deploys two OAIZIB-CM cartilage vulnerability models:

• —Radiographic OA prediction: KL >= 2
• —Advanced OA prediction: KL >= 3

Input

The app expects:

1. 1.A 3D NIfTI segmentation mask file: .nii or .nii.gz
2. 2.The segmentation must use OAIZIB-CM labels:
3. 3.1 = femur
4. 4.2 = femoral cartilage
5. 5.3 = tibia
6. 6.4 = medial tibial cartilage
7. 7.5 = lateral tibial cartilage
8. 8.Demographics:
9. 9.Age
10. 10.BMI
11. 11.Gender code, as in OAIZIB-CM metadata

The app does not segment raw MRI. It requires a precomputed segmentation mask.

Output

The app returns:

• —Predicted probability of KL >= 2
• —Predicted probability of KL >= 3
• —Risk interpretation
• —Cartilage feature table
• —3x3 topographic cartilage maps
• —Interactive 3D knee segmentation viewer

Local run

pip install -r requirements.txt
shiny run --reload app.py

Docker run

docker build -t cartilage-shiny .
docker run -p 7860:7860 cartilage-shiny

Then open:

<http://localhost:7860>

Recommended deployment

Use Hugging Face Spaces with the Docker SDK. This is a live Python/Shiny app, so a static host is not sufficient.