To access the session recording, navigate to the content tab and click the view video button. 

Session Schedule

Find a specific presentation in the course by navigating to the timestamp indicated below.

0:00:00
Health-oriented sleep staging (HOSS) with AI: Making sleep stages reflect health outcomes
Haoqi Sun (United States)

0:16:35
Continuous sleep depth index annotation with deep learning yields novel digital biomarkers for sleep health
Shenda Hong (China)

0:30:45
Deep learning and generative AI for automatic sleep monitoring and disease prediction
Wei Chen (Australia)

0:49:00
Accurately predicting mood episodes in mood disorder patients: Insights from wearable sleep and circadian rhythm data using machine learning
Jaekyoung Kim (Korea, Republic of)

1:04:00
Bridging sleep in clinic and at home: An AI-powered sleep foundation model for precision brain health
Yue Leng (United States)

1:26:00
Question and answer

Summary

Sleep plays a vital role in physical and mental health, serving as a critical window into overall well-being. Advances in artificial intelligence (AI) have enabled the development of new digital sleep biomarkers, enhancing our ability to predict, monitor, and manage a wide range of health outcomes. This symposium brings together leading researchers from diverse disciplines and international perspectives to discuss cutting-edge AI applications in sleep science. The talks will showcase innovative methodologies, novel biomarkers, and real-world applications, demonstrating how these advancements are transforming our understanding of sleep and its interconnectedness with health. Insights will benefit researchers, clinicians, and technologists across the globe.

Dr. Haoqi Sun will first introduce the concept of Health-Oriented Sleep Staging (HOSS), expanding traditional sleep staging to reflect specific health outcomes. Using deep learning to analyze polysomnography (PSG) data, the team identifies distinct states within each sleep stage, maximizing predictive accuracy for conditions including dementia, ischemic stroke, atrial fibrillation, myocardial infarction, type 2 diabetes, hypertension, and depression. This approach surpasses conventional sleep staging in predicting health outcomes, opening new pathways for understanding how sleep architecture relates to systemic health. Dr. Shenda Hong will present a novel Sleep Depth Index (SDI), developed using deep learning and large-scale PSG data. The SDI provides a continuous measure of sleep depth, strongly correlating with arousal and other nuanced sleep structures. This new digital biomarker overcomes traditional sleep staging limitations, revealing unique sleep subtypes associated with health risks. The public SDI API facilitates easy integration into diverse datasets, enabling researchers to uncover valuable insights into sleep and health. Dr. Wei Chen will discuss how deep learning and generative AI are revolutionizing automatic sleep monitoring and staging, addressing challenges such as data heterogeneity, imbalance, and heterogeneity. These AI-driven methods provide robust solutions for disease monitoring and prediction, focusing on sleep biomarkers linked to cardiovascular and Parkinson’s disease. The integration of AI into sleep research offers unprecedented opportunities to identify high-risk individuals and tailor interventions. Dr. Jae Kyoung Kim will demonstrate how wearable devices and machine learning enable accurate prediction of mood episodes in patients with mood disorders. By analyzing sleep-wake patterns and circadian rhythm data from 168 patients, the model achieves high predictive accuracy for depressive, manic, and hypomanic episodes. These findings highlight the potential for wearable technology to support personalized care and proactive management of mood disorders. Finally, Dr. Yue Leng will present an AI-powered sleep foundation model that bridges in-clinic and home-based sleep monitoring. This model leverages existing PSG data to develop a sleep foundation model and integrates it with home monitoring devices to provide novel insights into brain health. By enabling longitudinal tracking of sleep patterns, this approach supports early detection and management of diseases such as neurodegenerative diseases, REM sleep behavior disorder (RBD), and mood disorders.

This symposium highlights AI's transformative role in leveraging sleep as a lens for health, showcasing innovative frameworks and digital biomarkers that translate sleep research into actionable insights, advancing personalized medicine and broader health monitoring.

Learning Objectives:

Upon completion of this CME activity, participants will be able to:
• List health outcomes that could be reflected in sleep recordings and understand ways to extract health information from sleep using artificial intelligence
• Introduce a new concept of health-oriented sleep staging
• Recognize data-driven deep learning methods for AI-based digital biomarker
• Learn how to design deep learning objective functions to derive a novel Sleep Depth Index (SDI) from PSG data; Know how to utilize the public SDI API to calculate SDI in different datasets
• Identify the role of automatic sleep monitoring in advancing research and learn how deep learning and generative AI methods address challenges in sleep data analysis, including algorithm performance, robustness, and data heterogeneity and imbalance
• Evaluate methods for monitoring and prediction of cardiovascular disease and Parkinson's disease using sleep biomarkers
• Learn how wearable sleep-wake data can predict mood episodes and explore key steps in building sleep-based mood prediction models
• Recognize the advantages of streamlined, data-light prediction models
• Identify the significance and potential benefits of AI-powered sleep foundation models in advancing sleep health
• Explore the development process of sleep foundation models, from in-clinic polysomnography data to applications in home settings
• Assess the role of AI and home sleep monitoring devices in managing and improving brain health outcomes