In this investigation we develop and validate deep-learning models to predict patients’ risk of developing knee osteoarthritis and requiring total knee replacement within five years.

Osteoarthritis (OA) is a debilitating joint disease affecting millions worldwide. Early detection is crucial for effective management and to prevent severe joint damage. Our research focuses on developing cutting-edge tools to predict who is at high risk of developing and progressing with knee OA. By identifying these individuals sooner, we create a vital window of opportunity for early intervention and personalized treatment strategies.

To predict the likelihood of OA development and the potential need for total knee replacement, we are pioneering the use of deep learning models that analyze clinical risk factors and medical images, such as X-rays and MRIs. Our preliminary findings, based on studies with data from the Osteoarthritis Initiative and the Multi-Center Osteoarthritis Study, have shown significant promise. Our goals are to enhance the predictive accuracy, improve the generalizability and validate our models in order to deliver a widely accessible and automated risk assessment tool.

One key challenge in medical imaging analysis—and therefore also in this project—is the variability among scanners and patient populations. To overcome it, our team is developing novel deep-learning preprocessing pipelines and foundational models, then rigorously testing them on real-world patient data from multiple institutions.

Deep learning has the potential to revolutionize how we approach OA, enabling clinicians to identify at-risk patients early and improve the efficiency of clinical trials, leading in turn to faster development of new treatments. Our work is a critical step towards a future where the progression of this common and debilitating disease can be effectively slowed, perhaps even halted.