Â鶹ÍøÕ¾

Abstract

Disrupted sleep is increasingly recognized not only as a consequence but as an early sign of underlying health issues. It has been linked to the onset and progression of neurological and chronic conditions, including dementia and epilepsy. In this way, sleep serves as a vital indicator of brain health. To assess sleep, an intensive process is conducted inside specialized laboratories, involving the attachment of dozens of sensors to record physiological signals overnight. While effective, this method is costly,
limited to a single night's data, and often has long waitlists. Such inconvenient setup canbe intolerable for many populations, especially older adults or those with chronic illnesses. To improve sleep monitoring, wearables have become more widely adopted. These systems offer convenience and long-term tracking, but often lack the accuracy and resolution needed for clinical decision-making.

This talk explores how artificial intelligence—particularly signal processing, machine learning, and sensor fusion techniques—can enhance the diagnostic power of wearable sleep data. By leveraging algorithms to extract meaningful patterns from noisy, low-resolution signals, AI can transform consumer-grade wearables into tools capable of supporting medical-grade insights. This intersection is high interdisciplinary and presents a rich area of innovation: from hardware design and embedded sensing to data
analysis, model interpretability, and clinical validation.