AI Generates Future Knee X-Rays to Predict Osteoarthritis Progression Risk

Current diagnostic tools allow doctors to assess damage only after it occurs, making it difficult to predict how quickly a patient’s condition will worsen. Now, researchers have developed an artificial intelligence (AI) system that predicts what a patient’s knee X-ray will look like a year in the future, offering an unprecedented visual forecast of disease progression.

Developed by researchers at the University of Surrey (Guildford, Surrey, UK), the technology uses advanced machine learning to generate a realistic “future” X-ray image alongside a personalized risk score, helping doctors and patients visualize how osteoarthritis may evolve over a period of time. The system was trained on nearly 50,000 knee X-rays from about 5,000 patients—one of the largest osteoarthritis datasets ever assembled.

By combining this extensive dataset with innovative model design, the Surrey-developed AI not only surpasses comparable predictive tools in accuracy but also operates approximately nine times faster while maintaining a compact, efficient architecture. Researchers say this blend of speed, scale, and transparency positions the tool for seamless integration into clinical workflows.

Using a generative diffusion model, the AI creates a “future” version of the patient’s knee X-ray and highlights 16 key points within the joint. These markers indicate the specific regions monitored for structural change, allowing clinicians to interpret how and where deterioration is most likely to occur. This transparency makes the system more intuitive for medical professionals to trust and act upon.

The researchers believe the system could transform how osteoarthritis is monitored and managed. By showing both the current and projected X-rays side by side, doctors can more effectively communicate the urgency of intervention, while patients can better understand the impact of treatment adherence and lifestyle modifications. The technology may also pave the way for similar predictive models in chronic conditions such as lung disease and heart disease, providing the same level of visual insight and early warning.

“Earlier AI systems could estimate the risk of osteoarthritis progression, but they were often slow, opaque and limited to numbers rather than clear images,” said Gustavo Carneiro, CVSSP Professor of AI and Machine Learning. “Our approach takes a big step forward by generating realistic future X-rays quickly and by pinpointing the areas of the joint most likely to change. That extra visibility helps clinicians identify high-risk patients sooner and personalize their care in ways that were not previously practical.”

Related Links:
University of Surrey