Artificial intelligence analyzing ECG data during overnight sleep tests identifies future risk for heart failure and death.
RT’s Three Key Takeaways:
- Cardiac Risk Prediction: An artificial intelligence model analyzing single-lead heart monitor data during sleep studies can predict 10-year risk for atrial fibrillation, heart failure, and all-cause mortality, according to a study supported by the National Institutes of Health (NIH).
- Independent Predictive Value: The model maintained its predictive performance after adjusting for conventional cardiovascular risk factors, including diabetes and hypertension, as well as sleep-related metrics like sleep apnea severity, according to researchers.
- Accessible Data Collection: Because the method evaluates a single electrical signal rather than a standard 12-lead setup, the approach could enable streamlined data capture through low-cost patches, according to the study authors.
Electrocardiograms (ECGs) administered during overnight sleep studies and paired with sleep stage data can determine a patient’s risk for future adverse cardiac events, according to a research team supported by the National Institutes of Health (NIH). The findings, which evaluated a deep learning approach to assist clinical decision-making, were published in the journal Sleep.
Overnight sleep tests, known as polysomnography, serve as the standard method for diagnosing disorders such as obstructive sleep apnea and insomnia, both of which are established cardiovascular risk factors. Although ECGs are routinely recorded during polysomnography, the electrical recordings are rarely analyzed in depth, according to the study authors.
“This new approach has the potential to identify individuals at risk for cardiovascular disease years before clinical symptoms arise,” said David Goff, MD, PhD, acting director of NIH’s National Heart, Lung, and Blood Institute (NHLBI), in a news release. “Evaluating whether this information improves traditional risk prediction is an important next step. In the future, proven analytical approaches like this could be integrated into existing diagnostic tests for patients undergoing sleep studies as a screening tool for multiple adverse cardiovascular outcomes,” said Goff in a news release.
Model Training and Multi-Center Validation
To test whether routine overnight recordings could forecast 10-year risk for stroke, atrial fibrillation, myocardial infarction, heart failure, and all-cause mortality, researchers developed a deep learning architecture that combines single-lead ECG data with expert-annotated sleep stage information, according to the study.
The deep learning model was fine-tuned using a dataset of 15,809 patients from Massachusetts General Hospital in Boston, according to the researchers. Investigators subsequently evaluated performance across two separate patient cohorts: 9,810 patients from Emory University Hospital in Atlanta, and 12,576 patients from Beth Israel Deaconess Medical Center in Boston, with long-term clinical outcomes extracted from electronic healthcare records.
The analysis showed that the model successfully stratified patients into categories of long-term cardiovascular risk, with higher model scores correlating with higher adverse event rates, according to the study. The tool retained its predictive accuracy after adjustments for age, sex, body mass index (BMI), hypertension, diabetes, sleep apnea severity, sleep efficiency, and arousals. The system specifically predicted atrial fibrillation, heart failure, and all-cause mortality, according to the authors.
Clinical Implications for Sleep Medicine
“Importantly, we found that adding the score from our deep learning model improved prediction of cardiac outcomes beyond established risk factors,” said Gari Clifford, DPhil, 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, in a news release. “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 Clifford in a news release.
The research team noted that further refinement of the model is necessary to determine predictive capabilities for myocardial infarction and stroke. However, because the protocol requires only a single ECG lead rather than a traditional 12-lead arrangement, the authors reported that the technology could facilitate accessible, single-night data collection using low-cost sensor patches.
