An artificial intelligence algorithm, paired with smartwatch sensors, accurately diagnosed structural heart diseases in adults, including weakened pumping ability, damaged valves, and thickened heart muscle.
Researchers developed the AI tool using more than 266,000 12-lead electrocardiogram recordings from over 110,000 adults, and then tested it on single-lead ECGs captured by a smartwatch’s electrical heart sensor. The algorithm achieved an 88 per cent accuracy in detecting structural heart disease among 600 participants who wore smartwatches.
This is the first prospective study to show an AI algorithm can detect multiple structural heart diseases based on measures taken from a single-lead ECG sensor on the back and digital crown of a smartwatch, according to preliminary findings presented at the American Heart Association’s Scientific Sessions 2025.
“Millions of people wear smartwatches, and they are currently mainly used to detect heart rhythm problems such as atrial fibrillation,” said study author Arya Aminorroaya, an internal medicine resident at Yale New Haven Hospital and research affiliate at the Cardiovascular Data Science Lab at Yale School of Medicine. “Structural heart diseases, on the other hand, are usually found with an echocardiogram, an advanced ultrasound imaging test of the heart that requires special equipment and isn’t widely available for routine screening.”
Real-world conditions
The AI model maintained high performance when tested in real-world conditions. Using single-lead ECGs obtained from hospital equipment, the algorithm scored 92 per cent on a standard performance scale. The algorithm accurately identified most people with heart disease, achieving 86 per cent sensitivity, and was highly accurate in ruling out heart disease with 99 per cent negative predictive value.
Researchers validated the AI model using data from 44,591 adults seeking care at four community hospitals and 3,014 participants from the population-based ELSA-Brasil study in Brazil. To prepare the model for real-world use, researchers introduced interference, or “noise,” during training to help it handle imperfect signals from smartwatch sensors.
During the prospective study, 600 patients wore a smartwatch with single-lead ECG sensor for 30 seconds on the same day they received a heart ultrasound. The median age of the participants was 62 years, with approximately half of the participants being women. About five per cent were found to have structural heart disease on the heart ultrasound.
“On its own, a single-lead ECG is limited; it can’t replace a 12-lead ECG test available in health care settings,” said Rohan Khera, senior author and director of the CarDS Lab. “However, with AI, it becomes powerful enough to screen for important heart conditions. This could make early screening for structural heart disease possible on a large scale, using devices many people already own.”
Study limitations include a small number of patients with the actual disease in the prospective study and the number of false positive results. Researchers plan to evaluate the AI tool in broader settings and explore integration into community-based heart disease screening programmes.
