Overview
- University of Oxford researchers, whose peer-reviewed results appeared Wednesday in the Journal of the American College of Cardiology, trained and tested the tool on about 72,000 cardiac CT scans with British Heart Foundation funding.
- The AI reads subtle texture changes in fat surrounding the heart on CT images, a signal of underlying heart muscle damage that clinicians cannot see with standard imaging.
- In testing, the method predicted outcomes with roughly 86% accuracy, and patients in the highest-risk group were about 20 times more likely to develop heart failure within five years.
- Those flagged as highest risk faced about a one-in-four chance of heart failure over five years, giving doctors a clear window to step in with closer monitoring and targeted treatment.
- The team is working to adapt the approach to any routine chest CT and is exploring NHS rollout, a shift that could reach hundreds of thousands of patients each year and ease hospital pressure through earlier care.