Researchers at the University of Surrey have developed an AI system that can predict how a patient’s knee X-ray will look a year in the future, potentially transforming osteoarthritis diagnosis and care.
Presented at MICCAI 2025, the technology uses advanced machine learning and diffusion models to create a realistic “future” X-ray and provide a risk score showing how the disease may progress. This dual output helps doctors and patients visualize the likely development of osteoarthritis, making it easier to plan treatment and encourage preventive action.
The AI was trained on nearly 50,000 X-rays from about 5,000 patients, one of the world’s largest osteoarthritis datasets. It performs nine times faster than similar systems while being more compact and accurate, enabling faster use in real clinical environments.
Lead researcher David Butler said the tool bridges a key communication gap in medical AI by showing not just numbers but clear images of future joint health. Seeing both current and predicted X-rays side by side can motivate patients to follow treatment plans and help clinicians intervene earlier.
The model also identifies 16 key points in the knee joint, showing which areas are changing and increasing transparency for doctors.
According to Professor Gustavo Carneiro, this approach represents a major leap forward, offering both speed and interpretability. Beyond osteoarthritis, the team believes similar models could one day predict the progression of other chronic diseases such as lung damage or heart disease, giving healthcare professionals a powerful new way to act before conditions worsen.
The university is now exploring partnerships to bring the AI technology into real-world clinical use.
