AI Tool Predicts Long-Term Cardiovascular Risk from Sleep Study ECG

These studies routinely record electrocardiogram signals, but the data are rarely used to assess long-term cardiovascular risk. To make better use of these signals, researchers have developed a deep learning approach that analyzes sleep-study electrocardiograms to predict future cardiovascular outcomes.

A National Institutes of Health (NiH)-supported research team evaluated a deep learning model that analyzes single-lead electrocardiograms from overnight polysomnography. The model combines the electrocardiogram signal with expert-annotated sleep stage information. It was designed to estimate 10-year risk for cardiovascular outcomes, including atrial fibrillation, stroke, myocardial infarction, heart failure, and all-cause mortality.

The model was fine-tuned using data from 15,809 patients at Massachusetts General Hospital in Boston. Its performance was then assessed in two separate data sets that included 9,810 patients from Emory University Hospital in Atlanta and 12,576 patients from Beth Israel Deaconess Medical Center in Boston. The outcomes were derived from electronic health records.

The system separated patients into groups with different levels of long-term cardiovascular risk. Higher model scores were associated with higher risk. The model retained predictive value after adjustment for age, sex, body mass index, diabetes, hypertension, sleep apnea severity, arousals, and sleep efficiency.

The model predicted atrial fibrillation, heart failure, and death from any cause. The investigators reported that additional optimization is needed for myocardial infarction and stroke prediction. The findings were published in SLEEP on October 9, 2026, and the approach may be feasible because it uses a single electrocardiogram lead rather than a standard 12-lead setup.

“Importantly, we found that adding the score from our deep learning model improved prediction of cardiac outcomes beyond established risk factors. Adding these scores into sleep studies has great potential for improving early detection of cardiovascular disease, determining who is most at risk and being more proactive in patient care,” said Gari Clifford, D.Phil., study author, chair of the Department of Biomedical Informatics at Emory University School of Medicine and professor of biomedical engineering at Georgia Institute of Technology.

Related Links
National Institutes of Health