Heart scan videos better predicted one type of heart failure
An artificial intelligence model outperformed combined clinical information and standard scan measurements in 5,732 older adults followed for a median 10.6 years.
Our read
Biomarkers- Studied:
- Study in people
- Why it’s aging research:
- It bears on staying healthy in midlife and later life.
Editor of record: Boris Djordjevic · AgingResearch.News
The 5,732 adults without heart failure in the community-based Atherosclerosis Risk in Communities cohort averaged 75 years old; 59% were women. This observational study compared four prediction models using clinical information, standard heart ultrasound measurements or raw ultrasound videos. Researchers assessed performance by repeatedly training and testing the models on different portions of the data.
Over a median 10.6 years, 451 participants developed heart failure with preserved ejection fraction: the heart’s main pumping chamber expelled at least half its blood per beat. The artificial intelligence video model predicted this outcome better than standard scan measurements alone or combined with clinical information. It added no predictive information for heart failure with reduced ejection fraction, where that chamber expelled less than half its blood per beat. That form developed in 359 participants.
Why it matters
For heart failure in older adults, the authors suggested examining what the model focused on to investigate disease processes and identify new scan-based markers of risk.
Caveats
Testing within one cohort leaves performance in other populations uncertain. The conference abstract did not give numerical prediction scores, so the size of the improvement is unclear.
The paper
Univ of Texas Southwestern Med Cntr; UT Southwestern Med Cntr
AHA Scientific Sessions 2026 Abstract, Nov 2, 2026
Presented at AHA Scientific Sessions 2026, abstract 13745
Comments
Be the first to comment.