Electrooculogram Signal Analysis in Quantifying Sleep Apnea

Sleep apnea is a common sleep disorder that can induce negative effects on cognitive function and even cardiovascular health; therefore, timely diagnosis is crucial for effective treatment. The severity of sleep apnea is clinically evaluated using the Apnea-Hypopnea Index (AHI), which is derived from several signals collected during polysomnography (PSG). However, PSG is a complex and expensive process. Characterizing eye movements via electrooculography (EOG) during a PSG study could provide alternative measures for the AHI and quantify sleep apnea. To this effect, this study examined several EOG signals from the Sleep Heart Health Study, a public database, for feature extraction. The extracted EOG features were analyzed on a per-epoch and a per-subject level to train machine-learning models for sleep apnea severity classification. The classification was performed for two age groups (Age <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\leq 50$</tex> and Age <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\leq 70$</tex>), and the models classified sleep apnea severity as either Normal or Severe. The best classification performance was observed in the Age <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\leq 50$</tex> group, using only three features, with an accuracy of 7 0. 8 3 %. These findings suggest a relationship between EOG-based features and sleep apnea, which could lead to the development of screening tools and early detection of sleep apnea-related conditions in younger populations.