Advancements in artificial intelligence (AI) and wearable technology hold immense promise in transforming the landscape of sleep apnea management. Mount Sinai researchers, backed by a $4.1 million NIH grant, are pioneering AI-driven models to predict adverse outcomes in sleep apnea patients. By leveraging insights into physiological markers and employing machine learning techniques, their approach aims to surpass conventional diagnostic methods. Concurrently, wearable devices developed by other research teams offer accessible alternatives for detecting sleep apnea with high accuracy. These innovations signify a paradigm shift towards personalized, data-driven approaches in diagnosing and managing sleep disorders.
Mount Sinai’s dedicated team of researchers has been granted $4.1 million by the National Heart, Lung, and Blood Institute (NHLBI) at the esteemed National Institutes of Health (NIH). Their mission? To pioneer artificial intelligence (AI) models aimed at forecasting adverse outcomes in patients grappling with obstructive sleep apnea.
Sleep apnea, characterized by irregular breathing patterns and periodic airflow blockages during sleep, casts its shadow over nearly 39 million adults in the United States alone. Its repercussions extend beyond mere sleep disturbances, often culminating in severe health issues like hypertension, diabetes, and increased vulnerability to conditions like long COVID, along with elevated all-cause mortality rates.
Currently, the primary diagnostic tool for sleep apnea is the Apnea-Hypopnea Index (AHI), which tallies the frequency of apneas (breathing pauses) and hypopneas (partial breathing obstructions) during sleep. However, the AHI’s efficacy in predicting patient outcomes is limited, underscoring the pressing need for more refined diagnostic and prognostic instruments.
In response to this imperative, Mount Sinai’s research team is spearheading an AI-driven approach designed to gauge obstructive sleep apnea patients’ risk of adverse outcomes. Leveraging insights into the condition’s physiological underpinnings and its disruptive impact on vital sleep functions such as sleep stages, breathing patterns, and oxygen levels, the team aims to formulate a comprehensive risk assessment tool capable of predicting both short- and long-term outcomes, including excessive daytime sleepiness, neurocognitive impairments, and cardio-cerebrovascular morbidity.
To accomplish this ambitious goal, the researchers have crafted two machine learning (ML) models adept at forecasting patient risks by integrating data on ventilatory, hypoxic, and arousal variables. These models underwent rigorous training and were subsequently deployed to evaluate adverse outcome risks across a cohort of 11,000 sleep apnea patients.
The analysis yielded promising results, with one of the models boasting an impressive 87 percent accuracy in predicting apnea-induced sleepiness, compared to the AHI-based model’s meager 54 percent accuracy. Furthermore, in a separate cohort of 4,700 individuals, the AI-driven approach demonstrated over 80 percent accuracy in predicting cardiovascular mortality, far surpassing the standard method’s 58 percent accuracy rate.
Principal investigator Dr. Ankit Parekh, Director of the Sleep and Circadian Analysis (SCAN) Group and Assistant Professor of Medicine (Pulmonary, Critical Care, and Sleep Medicine) at the Icahn School of Medicine at Mount Sinai, expressed optimism about the project’s potential impact. He emphasized the utilization of cutting-edge AI models to profile sleep apnea patients based on data gleaned from routine sleep studies, a paradigm shift poised to redefine clinical management strategies for the condition.
Looking ahead, the research team plans to validate their models in a cohort of patients from the Mount Sinai Integrative Sleep Center. These participants will undergo polysomnogram sleep studies to capture comprehensive data on brain activity, oxygen saturation levels, heart rhythms, and breathing patterns during sleep. Subsequently, participants will be monitored and requested to maintain digital sleep diaries for three months post-study.
The ensuing findings will undergo rigorous retrospective validation and statistical scrutiny, paving the way for further advancements in sleep apnea management.
This groundbreaking endeavor represents the latest stride in leveraging digital health technologies to enhance outcomes for individuals grappling with sleep apnea. Last year, a team of researchers from the Georgia Institute of Technology (Georgia Tech) unveiled a wearable device capable of detecting sleep apnea with an impressive 88.5 percent accuracy.
The wearable device, adorned with forehead- and chin-mounted patches, offers a potential alternative to conventional polysomnography tests, renowned for their costliness and limited accessibility. By harnessing Bluetooth technology, the device captures and transmits signals reflecting brain, eye, and muscle activity, culminating in the generation of a comprehensive sleep score. Its remarkable accuracy underscores its potential utility as a valuable adjunct in sleep apnea detection, pending clinical validation.
The convergence of artificial intelligence (AI) and wearable technology holds immense promise in revolutionizing the management of sleep apnea. Mount Sinai’s innovative AI-driven approach, backed by substantial NIH funding, demonstrates significant potential in predicting adverse outcomes with unprecedented accuracy. By moving beyond traditional diagnostic methods, these advancements pave the way for personalized, data-driven approaches in sleep disorder management. Simultaneously, wearable devices developed by other research teams offer accessible alternatives for detecting sleep apnea, further enhancing diagnostic capabilities. As these technologies continue to evolve, they stand poised to transform the landscape of sleep medicine, ushering in an era of enhanced patient care and improved outcomes.
