Reimagining service delivery in the agentic era with Google Public Sector
cloud.google.com·21h ago
AI Research
arXiv CS.AI·Jun 9·5 min read
Knee-xRAI: An Explainable AI Framework for Automatic Kellgren-Lawrence Grading of Knee Osteoarthritis
TL;DR
Grading knee osteoarthritis (KOA) using the Kellgren-Lawrence (KL) scale is often inconsistent among different radiologists. Knee-xRAI is an explainable AI framework that automates this grading process by simulating clinical workflows.
✦ Why It Matters
Engineers can leverage explainable AI frameworks like Knee-xRAI to improve decision-making in medical imaging applications.
Key Takeaways
- Utilize Knee-xRAI for consistent KOA grading.
- Leverage deep learning models for enhanced diagnostic accuracy.
- Incorporate SHAP analysis for transparent decision-making.
- Adopt structured workflows to mimic clinical radiology processes.
How It Works
Knee-xRAI employs a multi-step deep learning pipeline that first segments joint space narrowing using a U-Net++ model. It then grades osteophytes at specific anatomical sites with an SE-ResNet-50 network and detects subchondral sclerosis through a hybrid texture-CNN.
The results from these analyses are combined into a 50-dimensional feature vector, which is evaluated by an XGBoost-SHAP classifier for auditing and a ConvNeXt hybrid predictor for deployment.