Random Forest Predicts Radiographic Knee OA Progression with AUC 0.87 Using Longitudinal MRI Features

Random forest using longitudinal quantitative MRI features predicted radiographic knee OA progression (AUC 0.87) vs baseline alone 0.80; composite progression AUC 0.66-0.70. Retrospective, no external validation.

• Retrospective analysis of 600 participants from the FNIH/OAI case-control cohort; data split 80/20 with 10-fold cross-validation.
• Key predictive features included medial femoral cartilage loss, bone marrow lesions in medial tibia and femur, medial osteophyte, and effusion-synovitis volumes.
• Single-cohort, case-control design without external validation limits generalizability.

RadPigeon summaries are original and for information only. They are not clinical advice.