Physiology Guided AI/ML Approach for Redefining Severity of OSA

2023 Focused Projects Grant for Junior Investigators 

Sajila Wickramaratne, PhD
Icahn School of Medicine at Mount Sinai

Key Project Outcome

Obstructive sleep apnea (OSA) affects millions of adults, yet the main tool physicians use to define its severity is the apnea-hypopnea index (AHI). The main drawback of using AHI is it is a poor predictor of short- and long-term outcomes of OSA. Our research proposes a new way forward: instead of AHI, we calculate the different types of physiological burdens based on the multi-dimensional nature of OSA. These physiological burdens are derived from Flow, SPO2, EEG and Heartrate signals from the polysomnography. By combining these signals with modern machine learning approach, we can create a more complete profile of each person’s sleep disorder. In early work, this approach predicted daytime sleepiness with far greater accuracy than AHI, and we believe it will also help forecast who is at greater long-term risk of death. Ultimately, this project could give patients and doctors a clearer picture of how obstructive sleep apnea is affecting their health and who may benefit most from treatment. Moving beyond AHI toward a multidimensional, physiology-based measure of OSA will bring us closer to personalized sleep medicine, where care is guided not just by the value of AHI, but by how other physiological burdens affect the patient as well.