The Sleep and Circadian Analysis Group

The Sleep and Circadian Analysis Group (SCAN) investigates how sleep and circadian processes influence health across metabolic, cardiovascular, and neurological domains. SCAN's pioneering use of machine learning and explainable AI to characterize and predict OSA outcomes places it at the cutting edge of sleep apnea research nationally. Led by Ankit Parekh, PhD, Medicine (Pulmonary, Critical Care and Sleep Medicine), and Artificial Intelligence and Human Health, Icahn School of Medicine at Mount Sinai, we serve as a central resource for analytic support, data processing, and methodological development in sleep and circadian research. We collaborate with academic institutions and industry partners to support basic, translational, and clinical studies.

Our work is grounded in the integration of large, multimodal datasets through state-of-the-art AI/ML models. Sleep studies generate rich physiological data, including brain activity, respiratory signals, oxygen levels, heart rate, and limb movements, often collected over extended periods. We combine this data with neuroimaging and behavioral measures to better understand sleep physiology and its impact on health and disease.

Our Approach: Multimodal Data and AI Driven Analysis

We apply artificial intelligence, machine learning, and pattern recognition methods to analyze complex sleep data. Our work focuses on identifying physiological patterns that are not easily captured through traditional visual scoring approaches. By leveraging high performance computing resources such as the Minerva supercomputer, we process high resolution sleep datasets efficiently and on a scale.

We develop automated pipelines that transform raw sleep recordings into quantitative metrics, enabling reproducible and scalable analysis across large cohorts. These approaches support more objective, data-driven characterization of sleep and circadian function.

Quantifying Sleep Physiology and Disease

We develop and apply automated methods to identify and quantify patterns in sleep EEG and physiological signals. This work improves how sleep disorders such as sleep apnea are defined, measured, and classified, including the development of metrics that better capture disease burden and heterogeneity.

Understanding Variability in Outcomes

Sleep disorders affect individuals differently. We investigate sources of variability in outcomes such as sleepiness, vigilance, memory, and treatment response, with the goal of identifying markers that explain these differences and inform individualized care.

Technology and Innovation in Sleep Measurement

We evaluate and validate sleep technologies, including wearable devices and home-based monitoring systems. Through collaborations with industry partners, we establish objective performance metrics and support independent validation of emerging tools.

Simulation and Training Tools

We are developing simulation platforms that replicate sleep study data, providing training environments for clinicians and researchers. These tools allow users to learn how to interpret sleep signals without the need for live recordings or laboratory resources.

Standardization and Reproducibility

We are actively involved in establishing standards for how sleep data are acquired, named, and analyzed. This includes developing consistent data structures, acquisition protocols, and analytical workflows. Our work also addresses the integration of subjective and objective measures of sleep, an area not fully standardized within existing clinical guidelines.

Collaboration and Impact

We collaborate with research groups across the United States and internationally, as well as with industry partners developing sleep technologies. By combining clinical expertise with advanced data analysis methods, our goal is to improve how sleep disorders are measured, understood, and treated, and to advance data driven approaches to sleep and circadian health.

Development and use of SLEEPTRONIX for ambulatory assessment of sleep, temperature, and cortisol

Funded by NHLBI - R21HL173733

PI: Parekh

The project aims to address limitations in the current assessment of obstructive sleep apnea (OSA), a common chronic disorder affecting more than one billion people worldwide. OSA severity is typically defined using the apnea hypopnea index (AHI), which measures the frequency of respiratory events during sleep, but growing evidence indicates that AHI has limited ability to predict clinical outcomes. To improve characterization of OSA severity, this work builds on the development of ventilatory burden, a fully automated breath by breath measure of abnormal breathing during sleep that captures respiratory dysfunction independent of hypoxia or arousal events. At the same time, many wearable sleep technologies attempt to estimate AHI but lack key physiological signals such as EEG, resulting in inaccurate estimates of total sleep time and an inability to characterize underlying ventilatory abnormalities. This project proposes to evaluate SLEEPTRONIX, a low profile, low cost, open-source wearable sensor system designed to resemble adhesive bandages and capable of recording EEG, EOG, airflow, and oxygen saturation. Using innovative algorithms, the study will determine whether SLEEPTRONIX can accurately estimate both AHI and ventilatory burden. In a cohort of newly diagnosed OSA patients and age and sex matched healthy controls, the primary goal is to develop and refine the system for continuous multi night monitoring of sleep and breathing patterns. The platform will also be expanded to measure circadian rhythm parameters using temperature and cortisol sensors enabled by emerging nanotechnologies. Together, this work aims to demonstrate that comprehensive characterization of OSA pathophysiology can be achieved using wearable technology, potentially improving diagnosis and clinical management of the disorder.

Enhancing Cardiovascular Risk Prediction and Treatment Response in Sleep Apnea Patients Using Advanced AI-based Transformer Models in Multi-Modal Sleep Data

Funded by NHLBI - R01HL175992

PI: Nadkarni (Parekh, CO-I)

The project aims to improve prediction of cardiovascular (CV) risk and treatment response in patients with obstructive sleep apnea (OSA), a common disorder associated with increased cardiovascular morbidity and mortality. Although OSA is typically diagnosed using the apnea hypopnea index (AHI) and treated with continuous positive airway pressure (CPAP), AHI does not reliably predict individual cardiovascular risk, and CPAP treatment has not consistently reduced cardiovascular events in clinical trials. One limitation is that polysomnograms collect a large amount of physiological data, including heart rate, muscle activity, and brain signals, yet most of this information is not incorporated into clinical decision making. This project proposes to apply transformer based neural networks to analyze the multimodal and longitudinal data captured during sleep studies to better predict cardiovascular risk and identify which patients are most likely to benefit from CPAP therapy. Using data from large epidemiological cohorts and randomized clinical trials comparing CPAP with usual care, the study will develop and validate predictive models and evaluate their performance relative to traditional clinical metrics such as AHI. By leveraging the full richness of polysomnography data with modern machine learning methods, this work aims to improve cardiovascular risk stratification in OSA and support more personalized treatment decisions.

Examining Unique Markers and Risk Factors for AD/ADRD in Foreign- and Native-born Older Chinese Americans: A Study of Social Determinants of Health, Sleep Patterns, Neuroimaging, and Plasma Biomarkers

Funded by NIA - R01AG095513

PI: Li (Parekh, CO-I)

The project aims to address the underrepresentation of Chinese Americans in Alzheimer’s disease and Alzheimer’s disease related dementias (AD/ADRD) research, despite this group being one of the fastest growing populations in the United States. Health outcomes in Chinese American older adults may vary by generation status and social determinants of health such as acculturation, language proficiency, health care access, stress, and social support, yet these factors and their relationship to AD/ADRD risk remain poorly understood. Emerging evidence also suggests that sleep disruption, which is common in older adults, may interact with these social and biological factors to influence disease risk. This study will examine how social determinants of health, sleep disturbance, and biological markers jointly contribute to AD/ADRD risk among older Chinese Americans, with particular attention to differences between foreign born and native-born individuals. The research will recruit 250 participants for baseline assessments and follow a subset longitudinally, using objective measures including at home sleep testing, cognitive evaluations, PET and MR imaging, and blood based proteomic and single cell multiomic biomarkers. In addition, integrative network biology and machine learning approaches will be used to combine clinical, imaging, and molecular data to identify highly predictive diagnostic and prognostic biomarkers. By examining these relationships in a well characterized cohort, the project seeks to improve understanding of AD/ADRD risk and health disparities in Chinese American older adults.

SCAN

Ankit Parekh, PhD, MS

Sajila Wickramaratne, PhD is an Associate Scientist whose work centers on applying machine learning and explainable AI to sleep medicine. She earned his Ph.D. in Electrical and Computer Engineering from the University of New Hampshire in 2022, after completing his MSc and BSc in Electrical Engineering at the University of Moratuwa, Sri Lanka. Following his doctorate, he served as a Postdoctoral Fellow in the Division of Pulmonary, Critical Care and Sleep Medicine at the Icahn School of Medicine at Mount Sinai, where she developed predictive models for outcomes of Obstructive Sleep Apnea, built generative AI tools to simulate polysomnography signals, and helped maintain an OMOP-standard sleep study database. Her research has produced numerous peer-reviewed journal articles and conference papers, several oral and poster presentations at venues like the American Thoracic Society and SLEEP conferences, and a competitive grant from the American Academy of Sleep Medicine, on which she served as Principal Investigator.

Li Zhou (Lily), PhD is a postdoctoral fellow in the Division of Pulmonary, Critical Care, and Sleep Medicine. Her research focuses on obstructive sleep apnea and REM sleep behavior disorder, particularly their relationships with neurodegenerative diseases. Her current work leverages machine learning and artificial intelligence methods to characterize disease features of obstructive sleep apnea and to predict clinical outcomes.

Pavel Boulgakov is a recent graduate of the University of Minnesota - Twin Cities, where he earned a Bachelor of Science in Computer Science and Mathematics in 2024. His experience spans machine learning research, virtual reality development, and administrative work: as a Research Intern in the Myers Lab, he built and benchmarked deep learning architectures to predict phenotype from genotype using Parkinson's disease genomic data on a high-performance computing cluster; as a Virtual Reality Software Developer at the Earl E. Bakken Medical Devices Center, he created interactive medical visualization software in Unity for use by clinicians and students on devices like the Meta Quest 2 and HoloLens. Beyond his formal roles, he has pursued a range of independent projects, including graph convolutional networks for molecular property prediction, phylogenetic tree construction from COVID-19 genomic sequences, reinforcement learning experiments in robotic control, and a drone delivery simulation built with C++ and TypeScript. His technical toolkit includes Python (PyTorch, scikit-learn, Pandas), C/C++, TypeScript, AWS, Docker, and Unity.

Catherine Kenney is a biostatistician who provides statistical support to the Sleep and Circadian Analysis Group. She splits her time between this group and the Li Lab, which studies Alzheimer’s disease and related dementias in older Asian American populations. The two groups are currently collaborating on a study examining how biological, social, and environmental factors, including sleep, may influence the risk of Alzheimer’s disease and related dementias in Chinese American populations. Her research interests include study design, longitudinal data analysis, and causal inference.

Ben Fox is a fourth year PhD candidate at the Icahn School of Medicine, advised by Girish Nadkarni, MD, MPH, and Ankit Parekh, PhD. His research focuses on building foundational transformer models using physiological signal data from sleep study and ICU monitoring data to estimate risk. Similar to large language models, such as ChatGPT, Ben builds representation models that learn features of multichannel signal data that can be used as input into downstream models to estimate risk outcomes, such as cardiovascular disease in sleep or hemodynamic monitoring in the ICU. 

Ke Xu: After graduating from the University of Wisconsin-Madison with degrees in neuroscience and psychology, Ke is pursuing a Master’s in Systems and Computational Biology at the Icahn School of Medicine. He is currently co-mentored by Ankit Parekh, PhD, and Yun Soung Kim, PhD, where his work focuses on sleep wearable design and machine learning model development related to sleep and daytime vigilance in patients with obstructive sleep apnea.

Valerie Builoff: Valerie is a PhD student in the Icahn School of Medicine at Mount Sinai Graduate School of Biomedical Sciences, where she is part of the AI and Emerging Technologies program. She is interested in integrating multimodal imaging and clinical data to better understand disease mechanisms and enable more precise, personalized care. Her current research at the lab lies at the intersection of neuroimaging, artificial intelligence, and sleep/circadian biology, with a focus on understanding how sleep disorders like obstructive sleep apnea affect brain function and long-term health. Prior to starting her PhD, Valerie worked in the AI in Medicine program at Cedars-Sinai Medical Center, where her research centered on developing and validating automated methods to analyze cardiac imaging data. She holds a BS in Physiological Science from UCLA.

Siyun Yang: Siyun is a PhD student in the Icahn School of Medicine at Mount Sinai Graduate School of Biomedical Sciences, where she is part of the AI and Emerging Technologies program. She is interested in developing state-of-the-art machine learning models to better understand consequences of sleep disorders. Her research will utilize multiple modalities of data including EHR, time-series from sleep studies and clinical data.

Yishan Lin (Susanna), MHS: Susanna is a Clinical Research Coordinator at the at the Icahn School of Medicine, where she supports studies in sleep research, respiratory physiology, clinical trials, and aging populations. She coordinates participant visits, manages data collection, and assists with regulatory documentation and overall study operations. She holds a Master of Health Science in Epidemiology from Johns Hopkins University. Her professional interests include clinical trial operations, longitudinal study design, patient engagement, and data management.

Sleep Medicine

David M Rapoport, MD
Indu Ayappa, PhD
Andrew Varga, MD, PhD
Korey Kam, PhD
Thomas M Tolbert, MD

Mount Sinai Collaborators

Clara Li, PhD
Jessica Spat-Lemus, PhD
Bin Zhang, PhD
Neomi A Shah, MD
Mayte Suarez-Farinas, PhD
Oren Cohen, MD
Girish N Nadkarni, MD, MPH
Vaishnavi Kundel, MD, MS
Jason C Kovacic, MD, PhD

Artificial Intelligence Can Drive Sleep Medicine (Artificial Intelligence Can Drive Sleep Medicine - PubMed)
Haoqi Sun, Ankit Parekh, Robert Joseph Thomas

Ventilatory Burden as a Measure of Obstructive Sleep Apnea Severity Is Predictive of Cardiovascular and All-Cause Mortality (Ventilatory Burden as a Measure of Obstructive Sleep Apnea Severity Is Predictive of Cardiovascular and All-Cause Mortality - PubMed)
Ankit Parekh, Korey Kam, Sajila Wickramaratne, Thomas M Tolbert, Andrew Varga, Ricardo Osorio, Monica Andersen, Luciana B M de Godoy, Luciana O Palombini, Sergio Tufik, Indu Ayappa, David M Rapoport

Hypoxic burden - definitions, pathophysiological concepts, methods of evaluation, and clinical relevance (https://pubmed.ncbi.nlm.nih.gov/39229876/)
Ankit Parekh

Full publication list can be found here: https://www.ncbi.nlm.nih.gov/myncbi/ankit.parekh.1/bibliography/public/