AI model combining MRI, biomarkers, and clinical data improves prediction of knee osteoarthritis progression, offering potential for earlier, personalized treatment.

An artificial intelligence-assisted model that integrates MRI radiomics, biochemical biomarkers, and clinical data shows promising results in predicting the near-term progression of knee osteoarthritis. Developed by Ting Wang of Chongqing Medical University and colleagues, the model was published in PLOS Medicine and represents a significant step toward more personalized and timely care for patients with the condition. Knee osteoarthritis, a degenerative joint disease characterized by the gradual breakdown of cartilage, affects around 303 million people globally and often leads to chronic pain and the need for joint replacement surgery. Early identification of patients at high risk of rapid disease progression could allow for earlier interventions, potentially slowing or preventing severe joint damage. Previous studies have suggested that combining multiple data sources—such as MRI scans, blood and urine biomarkers, and clinical assessments—can improve prediction accuracy. However, few models have successfully integrated all three data types in a unified framework. To address this gap, Wang and team analyzed data from 594 individuals with knee osteoarthritis enrolled in the Foundation for the National Institutes of Health Osteoarthritis Biomarkers Consortium. The dataset included 1,753 knee MRI scans collected over two years, along with biochemical test results and clinical information. Using AI and deep learning techniques, the researchers trained a model—called the Load-Bearing Tissue Radiomic plus Biochemical biomarker and Clinical variable Model (LBTRBC-M)—on half of the data. They then tested its performance on the remaining half. The model demonstrated strong accuracy in predicting whether patients would experience worsening pain, structural joint space narrowing, both, or no progression within the next two years. Notably, when seven resident physicians used the model to guide their clinical judgments, their predictive accuracy improved from 46.9% to 65.4%, highlighting the tool’s potential to support medical decision-making. The authors emphasize that while the results are encouraging, further refinement and validation in larger, more diverse patient populations are necessary before widespread clinical use. They also note the importance of interdisciplinary collaboration in developing such models. “This study shows that combining deep learning with longitudinal MRI radiomics and biochemical biomarkers significantly improves the prediction of knee osteoarthritis progression—potentially enabling earlier, more personalized intervention,” the researchers state. Co-author Professor Changhai Ding added that the work marks a meaningful advancement in applying AI to extract clinically relevant insights from complex musculoskeletal data, paving the way for more precise and proactive patient care.