Your dreams may not foretell future, but way that your heart falls asleep just might, revealing future health risks years before symptoms arise. Researchers have now shown that in early stages of sleep, as nervous system undergoes rapid changes, heart electrical activity can determine future health risks.
All that is needed is data from one electrocardiogram ECG lead, placed on chest during sleep, combined with artificial intelligence for prediction.
AI Can Predict Multiple Cardiovascular Outcomes
With that, artificial intelligence can predict someone risk of irregular heartbeat, stroke, heart failure, or death from any cause in next 10 years. Findings published in journal SLEEP, with support from US National Institutes of Health, offering promising new approach for early detection and proactive care.
Potential To Identify At-Risk Individuals Years Before Symptoms
This new approach has potential to identify individuals at risk for cardiovascular disease years before clinical symptoms arise, says David Goff, acting director NIH NHLBI.
Goff said evaluating whether this information improves traditional risk prediction is important next step, highlighting potential integration into existing sleep study diagnostic tests.
Research team led by scientists at Emory University first trained deep learning system on public health data of nearly 16000 patients.
Training And Testing On Large Datasets
These patients enrolled in sleep study at Massachusetts General Hospital in Boston, and as part of trial had heart activity measured during sleep by ECG. Once trained, AI system then put to test on two more datasets, including nearly 10000 patients from Emory University Hospital and 13000 from Beth Israel Deaconess.
In end, model able to sort participants into different levels of long-term cardiovascular risk, improving prediction beyond established risk factors significantly and promisingly.
Risk Increases Per Standard Deviation
After adjusting for conventional risk factors and sleep characteristics, AI system could predict 10-year risk of several cardiovascular outcomes with strong predictive value and accuracy. For each standard deviation increase in AI risk score, there was 103 to 172 percent higher risk of atrial fibrillation, which is irregular heartbeat.
There was also 19 to 40 percent higher risk of stroke, 69 to 133 percent higher risk of heart failure, and 65 to 82 percent higher risk of death. Only one group showed significantly increased risk of future myocardial infarction, or heart attack, indicating selective predictive strength for certain cardiovascular outcomes and conditions.
Improving Traditional Risk Models
Importantly, we found that adding score from deep learning model improved prediction of cardiac outcomes beyond established risk factors, says Gari Clifford, study author professor biomedical engineering Georgia Institute. Clifford said adding these scores into sleep studies has great potential for improving early detection of cardiovascular disease, determining who most at risk, being proactive.
Issues with sleep, like insomnia, known risk factors for cardiovascular disease, during polysomnography PSG test, gold standard diagnosing sleep disorders, heart often recorded as well. But unless data throws up serious red flag, findings rarely considered, all that subtle information on cardiac function could be going to waste previously.
Lightest Sleep Especially Indicative
Traditional models for cardiovascular risk usually consider ECG recordings when person awake or walking, but during sleep they may also record important fluctuations and dynamics. Using deep learning algorithm, researchers found ECG recordings show strong predictive value for cardiovascular health outcomes when combined with sleep stages even after adjusting factors.
That true even when adjusting for other cardiovascular risk factors, such as age, sex, diabetes, hypertension, smoking, highlighting independent predictive power of sleep ECG data. During lightest stage of sleep, for instance, changes to heart rate patterns and electrical conduction seem to be especially indicative of future health risks and outcomes.
These rapid autonomic fluctuations may amplify subtle abnormalities in heart-rate dynamics or conduction patterns, authors explain, making early sleep particularly valuable window for risk assessment.
Modest Absolute Increases But Meaningful Improvement
In fact, incorporating ECG-derived data from sleep improved current cardiovascular risk models by up to 6 percent for atrial fibrillation, up to 9 percent for heart failure. And up to 7 percent for mortality prediction, these absolute increases might seem modest, authors explain, but still improve ability to separate individuals who develop heart failure.
With continued development, AI system has potential to help health practitioners screen for patients with sleep disorders who may be at extra risk of heart failure.
In future, proven analytical approaches like this could be integrated into existing diagnostic tests for patients undergoing sleep studies as screening tool for multiple adverse outcomes, Goff said.
Q&A
Q: How can way you sleep reveal heart failure risk up to 10 years in advance using ECG and AI?
A: Researchers showed early sleep nervous system changes plus single-lead ECG chest data with deep learning AI can predict atrial fibrillation, stroke, heart failure, death risk 10 years.
Q: What did NIH NHLBI acting director David Goff say about potential of sleep ECG AI approach?
A: Goff said new approach has potential to identify individuals at risk years before symptoms arise, evaluating whether improves traditional risk prediction important next step, could integrate into sleep studies.
Q: How much did adding sleep ECG AI score improve prediction beyond traditional cardiovascular risk factors?
A: Study found per standard deviation increase, 103-172 percent higher atrial fibrillation risk, 69-133 percent heart failure, 65-82 percent death, improving models up to 9 percent for heart failure.
FAQ
Q: How was AI system trained and tested on sleep study datasets from Massachusetts General, Emory, Beth Israel hospitals?
A: Trained on nearly 16000 patients Massachusetts General Hospital sleep study with ECG, tested on nearly 10000 Emory University Hospital and 13000 Beth Israel Deaconess Medical Center Boston datasets.
Q: Why is lightest stage of sleep especially indicative of future cardiovascular health risks according to researchers?
A: During lightest sleep, changes to heart rate patterns and electrical conduction seem especially indicative, rapid autonomic fluctuations may amplify subtle abnormalities in heart-rate dynamics or conduction patterns.
Q: Are sleep issues like insomnia already known risk factors for cardiovascular disease before this AI study?
A: Yes, issues with sleep like insomnia known risk factors for cardiovascular disease, but subtle cardiac information during polysomnography gold standard test often going to waste unless serious red flag.
Q: Do traditional cardiovascular risk models usually consider ECG recordings during sleep or only when awake?
A: Traditional models usually consider ECG recordings when person awake or walking, but during sleep they may also record important fluctuations showing strong predictive value when combined with sleep stages.
Q: Could this proven analytical approach be integrated into existing diagnostic tests for patients undergoing sleep studies?
A: Yes, Goff said in future proven approaches like this could be integrated into existing diagnostic tests for patients undergoing sleep studies as screening tool for multiple adverse cardiovascular outcomes.
Disclaimer: This article provides educational information based on study published in journal SLEEP supported by US National Institutes of Health and does not constitute professional medical advice, diagnosis, or treatment.
