AI Tool Detects Hard to Identify Heart Dysfunction from Standard ECGs

The researchers tested two versions of the model. One used standard 12 lead ECGs, while the other used a single ECG lead similar to the measurement captured by some wearable devices.

Researchers at Wake Forest University School of Medicine have developed an artificial intelligence (AI) tool that can help clinicians identify signs of heart failure, including heart failure with preserved ejection fraction (HFpEF), a type that can be difficult to detect during routine care.

The study, published in the Journal of the American Heart Association, found that the AI model could identify three types of heart dysfunction from standard electrocardiogram (ECG) data: reduced ejection fraction (rEF), mildly reduced ejection fraction (mEF), and heart failure with preserved ejection fraction (HFpEF).

Ejection fraction refers to the percentage of blood pumped out by the heart's main pumping chamber with each heartbeat. In HFpEF, the heart pumps out a normal proportion of blood but does not fill or function normally.

The researchers developed the AI model using more than 1 million ECGs from Atrium Health Wake Forest Baptist. They then tested it on a separate dataset of more than 72,000 ECGs from the University of Tennessee Health Science Center to assess its performance in another patient population.

The researchers tested two versions of the model. One used standard 12 lead ECGs, while the other used a single ECG lead similar to the measurement captured by some wearable devices.

Both models performed similarly, with the 12 lead model particularly effective at distinguishing patients with reduced ejection fraction from those without it. Its performance was somewhat lower for the other two forms of heart dysfunction, although researchers said it remained potentially useful.

The single lead model performed nearly as well as the 12 lead model. According to the researchers, this suggests that the technology could eventually be adapted for wearable devices, although the model was not tested using data collected directly from wearables.

The model also demonstrated a strong ability to detect reduced ejection fraction in pediatric patients, performing as well as or better than previously studied models. Researchers noted that the pediatric group was relatively small.

The study also found that the model generalized well across different demographic populations.

Commenting on the findings, Oguz Akbilgic, Ph.D., corresponding author and professor of artificial intelligence in the Department of Cardiovascular Medicine at Wake Forest University School of Medicine, said, “This is a major step forward in how we can use everyday clinical tools to catch heart failure earlier. Our AI model can detect various types of heart dysfunction from a simple, single-lead ECG alone — the same lead configuration captured by many smartwatches and wearable ECG devices — suggesting the model could eventually be adapted for wearable-based screening.”

Akbilgic added, “Some of these conditions can progress without noticeable symptoms and may not be found until they become more severe. Our model helps fill that gap by identifying electrical patterns in the heart that humans can’t easily see so clinicians can decide when additional heart failure evaluation is needed.”

The research team is now piloting the model in a family medicine clinic at Atrium Health Wake Forest Baptist to assess how it performs when incorporated into clinical care.

“We’re testing the tool in a real-world health care setting to determine whether it can help clinicians identify patients who need additional evaluation and how it might affect care and resource use,” Akbilgic said.

The study was partially funded by the National Heart, Lung, and Blood Institute of the National Institutes of Health.

Stay tuned for more such updates on Digital Health News