Mount Sinai researchers have been awarded a five-year, $4.1 million grant from the National Heart, Lung, and Blood Institute at the National Institutes of Health to develop and study an artificial intelligence (AI)-powered model that predicts adverse outcomes of obstructive sleep apnea. 

The experts say their model will better reflect the underlying physiology of the condition and the ways it impairs sleep, improving patient care and treatment.

In response to an international call to better diagnose and manage sleep apnea beyond the standard scale, the apnea-hypopnea index, the researchers at Mount Sinai developed an AI-powered approach that examines the sleep functions apnea is known to impair—breathing, oxygen levels, and sleep stages—and combines these categories into a probability score that predicts the risk of short- and long-term outcomes of the disorder. 

“Our proposal uses a state-of-the-art artificial intelligence model that risk-profiles sleep apnea patients using data from routine sleep studies,” says principal investigator Ankit Parekh, PhD, director of the Sleep And Circadian Analysis Group and assistant professor of medicine at the Icahn School of Medicine at Mount Sinai, in a release. “Our study will assess the real-world performance of an AI approach and offer crucial evidence needed to translate metrics that go beyond the apnea-hypopnea in assessing severity of obstructive sleep apnea into clinical practice. Achieving this would leave us poised to shift the paradigm in clinical management of obstructive sleep apnea.”

The AI-powered method combines fully automated metrics across possibly independent ventilatory, hypoxic, or arousal categories with data-driven weights to determine risk of adverse outcomes. Mount Sinai experts say preliminary data from three cohorts of nearly 11,000 participants suggests the machine-learning model could predict the probability of sleepiness due to apnea with an accuracy of about 87%. In contrast, the model using the existing apnea-hypopnea index predicted sleepiness at about 54% precision.

Using data from a cohort of more than 4,700 participants, the machine-learned sleep apnea probability of cardiovascular disease could predict cardiovascular mortality with an accuracy of nearly 81%, compared to a regression model with the existing index that predicted cardiovascular death at about 58% accuracy.

The research team plans to test their two machine-learning models on a group of Mount Sinai Integrative Sleep Center patients who will undergo polysomnogram sleep studies that record brain waves, oxygen levels, heart rates, and breathing during sleep. The findings will be retrospectively validated against sleep data for statistical analysis. The patients will be monitored for three months while keeping digital sleep diaries as they progress through their clinical care.    

Photo 69191607 © One Photo |