A daytime obstructive sleep apnea severity assessment framework.

Abstract

Obstructive sleep apnea (OSA) is a prevalent sleep disorder characterized by repeated episodes of complete or partial blockage of the upper airway. These episodes can interfere with sound sleep and have fatal health consequences. They can also reduce the flow of oxygen to vital organs, like the brain, and lead to brain damage. In this paper, we leverage the correlation between brain damage and OSA to design a prediction framework that can assess the severity level of the OSA condition, as well as estimate the Apnea-Hypopnea Index (AHI) using only features obtained during wakefulness. Our daytime severity screening tool can enable the prioritization of patients for PSG studies based on the severity of their condition and can enable a more timely perioperative risk stratification. The proposed framework has a two-layered design that first classifies patients into coarse-grained OSA severity categories, before proceeding with the fine-grained AHI estimation within the classified category in the second layer. The performance of this framework was evaluated using the PSG's AHI scores as a gold standard. Using the proposed framework, patients can be classified into the correct severity group with 99.6% accuracy and their AHI can be estimated within an error of 4.5 events/hour, making the proposed system a promising reliable, daytime alternative for OSA severity screening.