Researchers at Imperial College London have developed an artificial intelligence model that reads a routine electrocardiogram in under two seconds to flag reduced heart pumping function and aortic valve disease, according to updates from the British Heart Foundation. The tool offers a potential secondary screening method for clinicians using diagnostic data already collected during standard patient care.
AI Model Detection Rates and Clinical Cohort Data
The AI model was developed using 10.6 million electrocardiograms and accompanying clinical reports, according to a September 2 announcement by the British Heart Foundation. Testing across two separate patient cohorts containing 5,442 and 61,520 patients yielded distinct performance metrics for the targeted cardiac conditions.
Detection rates for reduced heart pumping function reached 77% in the first cohort and 81% in the second cohort, according to British Heart Foundation data. For aortic stenosis—a narrowing of the aortic valve—detection reached 90% in the initial group and 80% in the larger cohort. These figures highlight peak results across different patient groups rather than a single uniform metric.
Did you know? Approximately one billion electrocardiograms are performed worldwide each year, according to British Heart Foundation estimates, creating a vast pool of existing data for potential AI screening layers.
Comparing 12-Lead and Single-Lead AI Models in Heart Failure Screening
Separate research published in the Journal of the American Heart Association and led by Wake Forest University School of Medicine demonstrates that artificial intelligence can also identify multiple heart failure types, including heart failure with preserved ejection fraction (HFpEF). According to corresponding author Oguz Akbilgic, Ph.D., the study evaluated models using both traditional 12-lead ECGs and single-lead configurations comparable to those captured by commercial wearable devices.
The Wake Forest study utilized more than 1 million ECGs from Atrium Health Wake Forest Baptist and tested the system on over 72,000 ECGs from the University of Tennessee Health Science Center. Researchers found that models using a single lead performed similarly to 12-lead configurations for specific classifications, suggesting potential future adaptability for smartwatch and wearable screening tools, though wearable-collected data was not directly tested in the study.
Clinical Integration and Upcoming NHS Testing Timelines
Clinical testing involving 590 National Health Service patients is currently underway across London and Bristol, as reported by the British Heart Foundation. Researchers forecast that routine NHS integration could be approximately two years away, though this remains a research ambition subject to formal regulatory approval and healthcare adoption.
Both the British Heart Foundation and study authors emphasize that the technology functions strictly as a supplementary tool rather than an independent diagnostic system. The model cannot independently confirm or exclude disease; its primary purpose is to flag patients who may require a confirmatory echocardiogram—an ultrasound examination of the heart.
Frequently Asked Questions
Can the AI model independently diagnose heart conditions?
No. According to the British Heart Foundation, the tool cannot independently confirm or exclude disease. It is designed to act as a secondary set of eyes to flag patients who need further investigation via an echocardiogram.
How long does the AI model take to analyze an electrocardiogram?
Researchers at Imperial College London report that their model reads a routine electrocardiogram in under two seconds.
Can wearable devices use this technology right now?
While researchers at Wake Forest University School of Medicine found that single-lead models performed similarly to 12-lead setups—potentially opening the door for smartwatch integration—the model has not yet been tested using data directly captured from wearable devices.
When will the tool be available in routine clinical care?
Current pilot testing is underway with NHS patients in London and Bristol. Researchers forecast that routine integration could be approximately two years away, pending formal approvals and further clinical validation.
Want to stay updated on the latest medical AI breakthroughs? Subscribe to our newsletter or explore our related coverage on healthcare technology innovations.