The practical promise is to learn more from a scan that already exists. Cardiac CT is ordered for specific clinical reasons, such as investigating chest pain or possible coronary disease. An AI system may be able to analyze additional image patterns without requiring another invasive procedure, but that does not make CT a general screening test for everyone.
What the Study Examined
A 2026 study in the Journal of the American College of Cardiology examined coronary CT angiography from 72,751 adults across nine UK centres. The researchers developed and externally tested a radiomic profile based on epicardial adipose tissue, the fat that sits around the heart.
Radiomics turns image texture, shape, and intensity into measurements that can be analyzed statistically. The model used these patterns to separate groups with higher and lower rates of future heart failure. That is population-level risk prediction, not certainty about what will happen to one person.
Why the Fat Around the Heart Matters
Epicardial fat is metabolically active and closely connected to the heart and coronary vessels. Changes in the heart's local biological environment may affect the tissue's composition. AI can quantify subtle image patterns that are difficult to capture through ordinary visual reading alone.
The study reported strong discrimination in both its development and external-test cohorts. External testing matters because a model can perform well on familiar data yet lose accuracy when it reaches a different hospital, scanner, image protocol, or patient population.
What a Risk Signal Could Change
If future prospective studies show that using the score improves care, it could help clinicians decide who may benefit from closer monitoring or a more focused review of established risk factors. The American Heart Association's prevention framework similarly emphasizes combining risk estimates with clinical history, biomarkers, imaging, and evidence-based management.
The AI result would be one input to a conversation. It would not replace symptoms, examination, blood pressure, diabetes status, kidney function, existing heart disease, or professional judgment. It also should not encourage someone to seek a CT scan without a suitable clinical reason, because CT uses ionizing radiation.
What Still Needs Proof
This was an observational prediction study. It did not prove that acting on the score prevents heart failure. Prospective trials are needed to test whether the information changes decisions and improves outcomes without causing avoidable anxiety, unnecessary testing, or treatment.
Regulatory review, transparent performance information, secure data handling, and monitoring across populations and clinical settings also matter. A useful medical AI tool needs a clearly defined intended use and evidence that its safety and effectiveness hold up in real-world care.
Sources and Further Reading
- JACC: Early Prediction of Heart Failure From Routine Cardiac CT Using Radiomic Phenotyping of Epicardial Fat
- University of Oxford: New AI tool can predict heart failure at least five years before it develops
- American Heart Association: Risk-Based Primary Prevention of Heart Failure
- U.S. FDA: Artificial Intelligence-Enabled Medical Devices
Medical disclaimer: This article is for educational purposes only and is not medical advice. It does not diagnose, treat, or recommend any medical test, device, medication, or intervention. Speak with a qualified healthcare professional for personal medical concerns.