Work

Predicting Knee Osteoarthritis Progression Using Explainable Machine Learning and Clinical Imaging Data

Year: 2025

Type: article

Abstract: BackgroundKnee osteoarthritis (KOA) progression is variable and often misaligned between structural imaging and patient-reported symptoms. Although quantitative imaging tools and artificial intelligence (AI) are increasingly applied in KOA research, most models remain opaque, limiting clinical insight into which imaging and clinical features drive predicted outcomes.ObjectiveTo evaluate explainable machine learning models for predicting KOA progression using quantitative MRI and clinical data, comparing baseline versus 24-month change features for two outcomes: a composite endpoint (radiographic plus pain progression) and a radiographic-only progression endpoint.MethodsWe analyzed 600 participants from the Foundation for the National Institutes of Health Osteoarthritis Biomarkers Consortium (FNIH), a nested case–control cohort in the Osteoarthritis Initiative (OAI). The participants were grouped as composite progressors (n=194), radiographic-only (n=103), pain-only (n=103), and non-progressors (n=200). Input features included age, sex, body mass index, Kellgren–Lawrence grade, joint space width, WOMAC pain score, and quantitative volume MRI measurements of KOA-related, including cartilage, bone marrow lesions, osteophytes, effusion-synovitis, and Hoffa’s synovitis, at baseline and as 24-month change. Five classifiers (random forest, XGBoost, logistic regression, decision tree, and multilayer perceptron) were trained using stratified 10-fold cross-validation. We applied multiple explainability methods, including Gini importance, SHAP values, regression coefficients, and permutation importance. Discriminative performance was measured by the area under the ROC curve (AUC).ResultsRadiographic progression was most accurately predicted using longitudinal change features (random forest AUC = 0.87). Baseline features alone also yielded strong performance (AUC = 0.80). Composite progression was more difficult to predict (AUCs = 0.66–0.70). Across models, key predictors included medial femoral cartilage loss, BMLs in the medial tibia and femur, medial osteophyte, and effusion-synovitis volumes.ConclusionsExplainable machine learning using quantitative MRI enables interpretable prediction of KOA progression, particularly for radiographic change. This is the first study in the FNIH/OAI cohort to integrate longitudinal quantitative MRI features with model-agnostic explanations across multiple classifiers.

Source: Osteoarthritis and Cartilage Open

DOI: 10.1016/j.ocarto.2026.100832