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
Morphological flow analysis for high loop gain in adults and kids
Robert Thomas (United States)

0:17:10
Centralness of respiratory events: A novel tool to guide non-PAP interventions in sleep disordered breathing
Scott Sands (United States)

0:29:20
Beyond Desaturations: Leveraging Pulse Oximetry for Comprehensive Diagnosis
Henri Korkalainen (Finland)

0:42:50
Endotypic traits characterizing obesity and sleep-related hypoventilation in patients with obstructive sleep apnea
Wan-Ju Cheng (Taiwan)

0:55:40
Cluster analysis in OSA: Opening the black box of AI with explainable modelling
Daniil Lisik (Sweden)

1:12:30
Question and answer

Summary

Artificial intelligence (AI) is at the forefront of innovation in sleep medicine, offering ground breaking tools for the precise diagnosis and management of sleep-disordered breathing (SDB). This symposium underscores how advanced technologies are transforming diagnostic paradigms and fostering personalized therapeutic interventions. The focus lies on recent advancements in signal analysis—from conventional nasal airflow measurements to oximetry-derived photoplethysmography signals—and their application to conditions such as obstructive sleep apnea (OSA) and obesity hypoventilation syndrome (OHS) in both adult and paediatric populations. This symposium aims to highlight the transformative potential of AI in redefining clinical practice in sleep medicine.

The session features five talks presented by a distinguished group of speakers: three established academics (RT, SS, HK), a mid-career clinician (WJC), and a junior PhD student (DL). The presentations will cover diverse yet interconnected aspects of AI-driven advancements in SDB diagnosis and treatment.
(1) Dr. Thomas will discuss how carotid body-mediated respiratory drive imposes unique oscillatory patterns that conventional scoring methods fail to fully capture. He will introduce respiratory self-similarity as a straightforward and intuitive metric for identifying high-loop gain, a significant driver of SDB. The presentation includes data from adult clinical practice and the paediatric CHAT study, emphasizing the clinical implications of this novel approach.
(2) Dr. Sands will present the concept of the "centralness" of respiratory events, a matrix that distinguishes between obstructive and central respiratory events. This approach offers critical insights into the underlying mechanisms of SDB, guiding non-PAP treatment options such as oral appliances, oxygen therapy, and pharmacological interventions. By broadening the spectrum of treatment strategies, this innovative framework facilitates tailored patient care.
(3) Focusing on pulse oximetry analysis, Dr. Korkalainen will explore how metrics such as oxygen variability, sleep patterns, and other key indicators enhance the diagnostic precision of OSA. This deeper understanding, particularly for patients at increased cardiovascular risk, underscores the expanded utility of pulse oximetry in guiding informed clinical decisions.
(4) Obesity is a risk factor for sleep-related and daytime hypoventilation. Based on findings from a prospective clinical cohort, Dr. Cheng will discuss endotypic traits in obesity OSA patients with sleep-related hypoventilation. The endotypic traits are derived from polysomnography sleep studies using an established model-based approaches. Her research highlights a non-linear relationship between body mass index and endotypic traits, as well as complex interactions between chemoreflex sensitivity and sleep-related hypoventilation.
(5) Finally, Dr. Lisik will present work on explainable AI models, emphasizing the importance of transparency and clinical relevance. His talk will address the limitations of deep learning and advocate for statistical modelling as a means of creating actionable AI tools. Large-scale clinical applications will be discussed, showcasing how explainable AI can uncover novel insights into the mechanisms underlying SDB.

Through these presentations, the symposium provides an overview of the cutting-edge advancements reshaping SDB diagnostics and management. Attendees will gain a deeper understanding of the need to move beyond traditional methodologies, embracing AI-driven approaches to enable more precise, personalized, and effective patient care.

Learning objectives

Upon completion of this activity, participants will be able to:

· Explore carotid body-driven respiratory oscillations and the limitations of conventional scoring in detecting high loop gain in sleep apnea. They will be introduced to respiratory self-similarity as a diagnostic tool. The CHAT study will be used to illustrate its application in identifying high loop gain.
·   Assess the centralness of respiratory events.
· Recognize the association between upper airway obstruction and the "centralness" of events.
· Examine clinical indications for using event "centralness" to predict the treatment response of oral appliance, oxygen, and drugs.
· Utilize pulse oximetry beyond desaturation events to assess various aspects of sleep health. They will explore how pulse oximeters can provide valuable insights into oxygen variability, sleep patterns, and other critical metrics, enabling more comprehensive monitoring and enhancing diagnostic accuracy for sleep-disordered breathing.
· Investigate non-linear association between body mass index and endotypic traits. Obesity is a risk factor for sleep-related and daytime hypoventilation. While chemoreflex sensitivity is believed to be blunted in patients with obesity-related hypoventilation, this is not necessarily the case in sleep-related hypoventilation.
 
· Explore the concept of deep learning and its limitations. They will be introduced to sophisticated statistical modelling for developing clinically relevant AI models. Cluster analysis of sleep apnea phenotypes will be used as an example of AI implementation in large clinical settings.