Our understanding of the impact of sleep upon a broad range of health conditions is growing rapidly—from weight gain to cardiovascular health. The Sleep and Circadian Analysis Group provides analytic and data services, as well as an integrated knowledge base in support of research and clinical projects that investigate sleep and circadian behavior. As part of the Division of Pulmonary Critical Care and Sleep Medicine, we work with researchers conducting basic and translational research, as well as industry, non-profit organizations, and other research institutions.
Big Data Is Expanding Our Understanding of Sleep
There is a vast amount of data that can be gathered during sleep studies—or polysomnography—including brain waves, oxygen levels, heart rate, breathing, pupillometry, melatonin levels, as well as eye and limb movements. Magnetic resonance imaging (MRI), both structural and functional, can also yield insights on brain behavior and abnormalities during sleep. Under the leadership of Ankit Parekh, PhD, the Sleep and Circadian Analysis Group assesses different patterns using tools such as Pattern Recognition, artificial intelligence, machine learning, and other advanced analytical methods for multi-modal data. We are also leaders in establishing standard protocols for the acquisition, naming, and analysis of the data.
Over recent years, as data storage has become inexpensive, we are able to record sleep studies at a much higher resolution than ever before. With the power of Mount Sinai’s supercomputer, Minerva, and the assistance of the Scientific Computing group, we are able to analyze the data in ways and at speeds we were never able to achieve previously. We are able to put this processing power and expertise at the service of researchers around the world, in New York City, and within the Mount Sinai community. For example, we have collaborations with the University of Sydney in Australia, Stanford University, the University of Pittsburgh, the University of Wisconsin, and New York University. We also collaborate with industry, including manufacturers of sleep wearables and temperature-controlled mattresses.
Exploring the Many Facets of Sleep
Pattern Recognition of Sleep EEG: To better understand consequences of sleep disorders, we have developed and published automated approaches that quantify abnormal patterns in sleep EEG. Our automated approaches are publicly available and have been used by over 35 groups worldwide.
Burden of Sleep Apnea: We have developed methodology that enables better definition of disease burden in sleep apnea. Effective treatment of sleep apnea requires establishing the presence, severity, and type of sleep apnea. Our group has developed and published on automated approaches to this end.
Sleep Simulator: A sleep study generates an enormous amount of data: eight hours of brain activity, respiration, limb movements, and many other metrics. We are in the process of developing a sleep simulator for fellows and technicians to learn on, which enables them to understand the signals better. The simulator eliminates the need for subjects, lab time, data storage, and provides an optimal training environment to develop skills and expertise.
Eight Sleep Temperature-Controlled Mattresses: Eight Sleep is a manufacturer that makes products to monitor sleep non-invasively. However, the gold standard of sleep studies is polysomnography conducted in a sleep lab. In order to validate their technology, Eight Sleep approached us for independent validation.
Sleep Wearables: There has been an explosion of sleep wearables for consumer wellness, as well as medical use. We have collaborated with manufacturers to establish metrics to determine effectiveness. It enables them to outsource programming, statistical experts, data analysis—and also provide independent validation.
Quantifying the Patterns of Sleep
Standardized Data Recommendations: While the data used by sleep labs is somewhat standardized, naming conventions within the file formats may vary from institution to institution. In addition, the sampling frequency for each data point is not standardized. We publish recommendations for naming, acquisition, and processing. Part of the challenge is that for a condition such as sleep apnea, we are not just quantifying the disease burden, but also subjective outcomes. For instance, how does the patient feel the next day? And while the Academy of Sleep Medicine has some recommendations, there is no guidance on how to process that data to arrive at that metric. Our recommendations include mathematical models, programming language, and applications for AI and machine learning.
Development of Pattern Recognition Methods: Pattern Recognition methods are mathematical processes of quantifying the different patterns patients exhibit during sleep studies. Even today, physicians and technicians visually examine the data of an eight-hour sleep study, 30 to 60 seconds at a time. After visually identifying patterns that appear to represent disease or health through abnormal or normal patterns, the physician or technician will attempt to quantify through clinical metrics whether the individual is in need of treatment. We use years of data and experience in the clinical domain to develop mathematical models and other algorithms that can automatically take the data of an eight-hour study and generate a quantified numerical value for what would otherwise be a time-consuming endeavor.
Other Services
The Sleep and Circadian Analysis Group is a resource for researchers and sleep technicians inside and outside the Mount Sinai community including:
- Basic training and certification in actigraphy, polysomnography, and circadian analyses
- Assistance for investigators of all levels from junior to senior, in the interpretation and analysis of sleep and circadian data
- Assistant to internal and external researchers for High Performance Computing analysis of sleep and circadian data
- Assistance in the development of grant proposals aimed at utilizing sleep and circadian data