A heart ultrasound taken between 2002 and 2004 may have been quietly keeping a long-term score. When a deep learning model re-read the echocardiography clips from 568 adults, the people whose hearts looked older than their birth certificates were more likely to die of cardiovascular disease or dementia over the next 20 years.
Read next on Morning Overview: New FCC rules on how people can stop robocalls and texts take effect Nov. 25
Dr. Jeffrey G. Malins of Mayo Clinic in Rochester, Minnesota, presented the result on Nov. 13 at the European Society of Cardiology’s Digital and AI Summit in Basel, Switzerland. For every five years that a person’s AI-estimated heart age ran ahead of their calendar age, the study found a 61% higher risk of that combined death outcome. The work is a conference presentation, and the release describing it does not mention a published journal paper.
A heart-age gap measured from ultrasound clips
The European Society of Cardiology’s release on EurekAlert describes a retrospective analysis of a community-based group of adults aged 50 or older who had echocardiography in 2002 to 2004. Malins’s team applied a deep learning age model that had been developed earlier and averaged its predicted age across each person’s clips. The “cardiac age gap” is that AI-predicted biological age minus the person’s chronological age.
Earlier work had mostly tied an older AI-estimated heart age to worse outcomes among people who already had cardiovascular disease. According to News-Medical’s account, this study turns to the general population, where that evidence was thin. A separate ESC release on electrocardiogram-based heart age, presented at the EHRA 2025 congress, had reported that a heart age seven years above chronological age was linked to a 62% higher all-cause mortality risk in a far larger group trained on 425,051 ECGs.
61 percent, then 46 percent after more adjustment
The central figure is a hazard ratio of 1.61 (95% confidence interval 1.24 to 2.07) after adjusting for age and sex. Each additional five years of cardiac age gap went with a 61% higher chance of cardiovascular or dementia-related death within 20 years.
Adding body mass index and left ventricular ejection fraction, a standard measure of how much blood the heart’s main chamber pushes out, weakened the link without erasing it. The adjusted hazard ratio fell to 1.46 (1.13 to 1.90), a 46% higher risk per five-year gap, with a p-value of 0.004. That the association survives a correction for ejection fraction suggests the model reads something in the images beyond pumping strength, though the release does not say what.
A hazard ratio compares how fast events occur in one group against another over the follow-up period, and the figure is quoted per five-year step in the gap. If the relationship scales smoothly, which the release does not claim, a ten-year gap would imply a hazard ratio near 2.6, since 1.61 multiplied by itself is about 2.59. That extrapolation is arithmetic and not a reported result. The confidence interval of 1.24 to 2.07 already shows how wide the plausible range is for a cohort of 568 people.
The 20-year window matters as much as the size of the group. Because the echocardiograms date from 2002 to 2004, deaths could be tracked for two full decades, long enough for dementia-related mortality to accumulate in adults who were at least 50 when scanned. Imaging methods and treatments have changed since then, which is one reason the authors want newer, independent cohorts.
Why cardiovascular and dementia deaths were counted together
The outcome merges two causes of death, and Malins acknowledged that choice. He said the outcomes were combined for practical reasons and that separate analyses would help clarify how cardiovascular and neurodegenerative disease are connected. The result therefore cannot say whether an older-looking heart predicts dementia deaths on their own, which is how the finding is sometimes summarized.
His interpretation was cautious. He said AI-derived cardiac aging signatures may serve as a marker to identify people with long-term vulnerability. The release adds that the design is retrospective and observational, so it shows association and not cause, and it gives no count of deaths, no breakdown by sex or age, and no funding statement.
What larger studies would need to show
The authors call for prospective evaluations in larger, independent cohorts before anyone treats the measure as a screening tool. Two questions are named in the release: whether AI heart-age estimates can reliably identify high-risk individuals or track the effect of treatment, and whether lifestyle or medical interventions can narrow the cardiac age gap.
Coverage of the presentation, including Medical Xpress, repeats the same figures. For readers of the underlying science, the method sits in a lineage that includes electrocardiogram-based heart age estimation by a deep learning model, which looked at the incidence of cardiovascular disorders. What remains unsettled is whether a 568-person cohort, scanned two decades ago, will hold its 1.61 and 1.46 when repeated in a larger group.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
More from Morning Overview
- Cadillac cut the price of its cheapest EV, which goes up to 317 miles on a charge
- Winter weather advisories cover I-80, I-70 and the Bighorns with up to 15 inches of snow
- Arizona’s governor declared a statewide emergency after Flagstaff logged its wettest Nov. 1 on record
- Zero Motorcycles unveiled an electric motorcycle targeted at under $8K and a street-legal dirt bike at EICMA