Summary
Researchers at the University of Oxford, funded by the British Heart Foundation, have developed an AI tool that detects fat deposits around the heart in CT scans — invisible to human eyes — with 86% accuracy in predicting heart failure up to five years before it develops. The study analyzed 72,000 patients and was published in the Journal of the American College of Cardiology. The tool is now being evaluated for NHS-wide rollout and may be extended beyond cardiac CTs to any chest CT scan.
Key Points
- 86% accuracy rate in identifying at-risk patients from CT scan texture analysis.
- Patients in the highest-risk group were 20 times more likely to develop heart failure than low-risk patients, with a 1-in-4 chance of heart failure within five years.
- Heart failure is the second biggest killer in the UK — 54,000 deaths in 2024, 9.6% of all UK deaths.
- Currently ~350,000 patients are referred for cardiac CT scans in the NHS annually; rollout could screen this entire population.
- Lead researcher Prof. Charalambos Antoniades is working to extend the method to any chest CT, not just cardiac-specific scans — dramatically expanding potential reach.
- The UK government’s 10-year NHS plan (“Fit for the Future”) explicitly aims to make the NHS “the most AI-enabled health system in the world.”
Newsletter Angles
- The gap between AI capability and deployment: the tool exists, has 86% accuracy, has been validated on 72,000 patients — but it’s still in NHS rollout evaluation. Why does this always take so long? What are the structural barriers?
- This is the “AI that saves lives vs. AI that writes emails” contrast — a useful framing for any piece on what AI is actually good for vs. hype.
- The extension to any chest CT is potentially huge: millions of CT scans are taken for unrelated reasons; this tool could turn every chest scan into a free cardiac screening.
- NHS as AI testbed: a single-payer system is uniquely positioned to deploy and evaluate this kind of tool at population scale. What does it mean for the U.S. fragmented insurance model?
Entities Mentioned
- University of Oxford — developed the AI tool through Prof. Antoniades’ lab
- British Heart Foundation — funded the research; Dr. Sonya Babu-Narayan is clinical director
- NHS — target deployment system; 350,000 annual cardiac CT referrals
Concepts Mentioned
- AI in Healthcare — diagnostic AI reading textural changes invisible to human imaging
- Predictive Medicine — five-year pre-symptomatic risk identification
Quotes
“Although this study used cardiac CT scans, we are now working towards applying this method to any CT scan of the chest, performed for any reason. This will allow doctors to make more informed decisions about the best way to treat patients.” — Prof. Charalambos Antoniades
“Early heart failure diagnosis is crucial — it means doctors can better manage someone’s condition, which gives them a fighting chance of living longer in better health.” — Dr. Sonya Babu-Narayan, BHF
Notes
- Source is The Independent (UK), published April 8, 2026. Reputable outlet for NHS/UK health coverage.
- The 86% accuracy figure is from the study itself (Journal of the American College of Cardiology). No independent replication cited.
- The article does not mention false positive rate — a key missing metric for clinical deployment decisions.
- Microbira (CEO Marianne Ismail) is briefly quoted but is a company with a commercial interest in AI health diagnostics; treat as advocacy, not neutral commentary.