AI heart‑fat boosts CVD prediction
Mayo Clinic research shows adding AI‑derived heart‑fat measurements from routine imaging significantly improves long‑term cardiovascular disease risk prediction for patients with metabolic disorders, including diabetes. (newswise.com)
The manuscript, titled “Deep Learning–Derived Pericardial Adipose Tissue by ECG‑Gated Computed Tomography Predicts Cardiovascular Events Beyond Coronary Calcium,” was presented at the 2026 American College of Cardiology Scientific Session and published in the American Journal of Preventive Cardiology. Investigators applied the algorithm to nearly 12,000 adults and tracked outcomes over roughly 16 years to assess long‑term cardiovascular risk. The team used deep‑learning segmentation on standard, ECG‑gated coronary artery calcium CT scans to quantify pericardial (heart) fat volume automatically, with Zahra Esmaeili named as first author and Mayo Clinic’s preventive cardiology group listed among the authors. About 10% of study participants developed cardiovascular disease during follow‑up, and higher pericardial fat volume remained independently associated with elevated event risk after adjustment for traditional risk factors and coronary calcium scores. Adding the AI‑derived fat metric to models that included the AHA PREVENT equation and the coronary artery calcium score increased overall predictive accuracy, with the largest gains seen in people classified as low, borderline or intermediate risk by existing tools. Mayo Clinic authors, including senior author Francisco Lopez‑Jimenez, emphasize that the measure can be extracted from the same CT scan without additional testing or cost, positioning the tool as a potentially scalable way to refine prevention strategies pending broader validation.
