IPM Take
The sharp signal is opportunistic prevention. The tool does not ask health systems to create a new screening programme from scratch. It asks whether scans already being performed can reveal hidden risk before symptoms appear. That is powerful, but it also creates a policy question: if AI flags a person as high risk for heart failure, who follows up, what intervention is offered, and how is that pathway funded?
Executive Summary
A 2026 Journal of the American College of Cardiology study developed an AI approach to predict future heart failure risk from routine cardiac CT scans using radiomic phenotyping of epicardial fat. The University of Oxford reports that the tool was trained and validated in more than 70,000 individuals across nine NHS Trusts, with decade-long follow-up after cardiac CT scans. It was then tested in 13,424 people in England and predicted five-year heart failure risk with 86 percent reported accuracy. Oxford says the tool identifies textural changes in fat around the heart that are not visible to the human eye, and that researchers are seeking regulatory approval for NHS rollout.
Why it matters
- Clinicians: Could gain an earlier risk signal from scans already being performed, but need clear follow-up protocols before acting on AI-generated risk scores.
- Hospitals / providers: Would need regulatory approval, radiology workflow integration, data governance and referral pathways before routine deployment.
- Payers / public authorities: Must decide whether AI-based opportunistic prevention creates value by preventing heart failure or only adds monitoring burden.
Previously, routine cardiac CT scans were mainly used to investigate chest pain and assess coronary artery disease. The new AI approach adds a prevention layer by extracting risk information from tissue patterns that clinicians would not normally see.
What has changed is the use of epicardial fat radiomics as an early risk signal. The Oxford team reports that textural changes in fat around the heart may indicate inflammation and unhealthy heart muscle before heart failure develops. People in the highest risk group were reported to be 20 times more likely to develop heart failure than those in the lowest risk group, with around a one in four chance of developing the condition within five years.
The affected population is people already undergoing cardiac CT, not the general public. Oxford researchers are also working to adapt the tool to other chest CT scans, which could expand opportunistic risk detection beyond cardiac CT.
For IPM, the implementation issue is direct: prediction is only useful if it triggers action. Health systems will need regulatory approval, clinician workflows, risk communication, follow-up protocols and prevention pathways before AI-based heart failure prediction can improve outcomes at scale.
