Knee-xRAI: An Explainable AI Framework for Automatic Kellgren-Lawrence Grading of Knee Osteoarthritis

arXiv cs.CV / 4/28/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

Knee-xRAI addresses limitations in Kellgren-Lawrence (KL) grading for knee osteoarthritis by improving explainability and reducing inter-reader variability in radiographic assessment.
The framework independently measures the three key radiographic KOA features—joint space narrowing (JSN) via U-Net++, osteophytes via an SE-ResNet-50, and subchondral sclerosis via a hybrid texture-CNN—and then integrates them into KL grade classification.
It uses two complementary classification routes: an XGBoost path that enables SHAP-based feature attribution, and a ConvNeXt hybrid path that combines structured features with full-image encoding for higher accuracy.
On 8,260 OAI-derived radiographs, JSN segmentation reached a Dice coefficient of 0.8909 with strong agreement on mJSW (ICC 0.8674), while the ConvNeXt hybrid achieved test QWK 0.8436 and AUC 0.9017.
Ablation studies indicate JSN is the dominant predictor for KL grading, osteophytes provide incremental improvement, and sclerosis adds only marginally, with inference-time perturbations confirming the structured feature pathway’s contribution.

Abstract

Radiographic grading of knee osteoarthritis (KOA) with the Kellgren-Lawrence (KL) system is limited by inter-reader variability and the opacity of current deep learning approaches, which predict KL grades directly from images without decomposing structural features. We present Knee-xRAI, a modular framework that independently quantifies the three cardinal radiographic features of KOA (joint space narrowing [JSN], osteophytes, and subchondral sclerosis) and integrates them into an explainable KL grade classification. The pipeline combines U-Net++ segmentation for contour-based JSN measurement, an SE-ResNet-50 network for per-site osteophyte grading (OARSI scale), and a hybrid texture-CNN classifier for binary sclerosis quantification. The resulting 50-dimensional structured feature vector feeds two complementary classification paths. An XGBoost path supports SHAP-based feature attribution. A ConvNeXt hybrid path combines the structured vector with a full-image encoder for enhanced predictive performance. Evaluated on 8,260 radiographs from an OAI-derived dataset, the JSN module achieved a Dice coefficient of 0.8909 and an mJSW intraclass correlation of 0.8674 against manual annotations. The ConvNeXt hybrid path reached a test quadratic weighted kappa (QWK) of 0.8436 and AUC of 0.9017. The transparent XGBoost path achieved a test QWK of 0.6294 with full feature-level audit capability. Ablation confirmed JSN as the dominant predictor (QWK = 0.6103 alone), with osteophyte features providing consistent incremental gain (+0.0183) and sclerosis contributing marginally. Inference-time ablation of Path B confirmed the structured pathway contributes materially beyond the image encoder, with QWK drops of 0.098 (feature zeroing) and 0.284 (feature-image permutation). Knee-xRAI explicitly quantifies all three KL-defining radiographic features within a single auditable pipeline.