Apnea-hypopnea index estimation using quantitative analysis of sleep macrostructure

Obstructive sleep apnea, characterized by recurrent cessation or substantial reduction in breathing during sleep, is a prevalent and serious medical condition. Although a significant relationship between obstructive sleep apnea and sleep macrostructure has been revealed in several studies, useful applications of this relationship have been limited. The aim of this study was to suggest a novel approach using quantitative analysis of sleep macrostructure to estimate the apnea-hypopnea index, which is commonly used to assess obstructive sleep apnea. Without being bound by conventional sleep macrostructure parameters, various new sleep macrostructure parameters were extracted from the polysomnographic recordings of 132 subjects. These recordings were split into training and validation sets, each with 66 recordings including 48 recordings with an apnea-hypopnea index greater than 5 events h(-1). The nonlinear regression analysis, performed using the percentage transition probability from non-rapid eye movement sleep stage 2 to stage 1, was most effective in estimating the apnea-hypopnea index. Between the apnea-hypopnea index estimates and the reference values reported from polysomnography, a root mean square error of 7.30 events h(-1) was obtained in the validation set. At an apnea-hypopnea index cut-off of  ⩾30 events h(-1), the obstructive sleep apnea diagnostic performance was provided with a sensitivity of 90.0%, a specificity of 93.5%, and an accuracy of 92.4% by our method. The developed apnea-hypopnea index estimation model has the potential to be utilized in circumstances in which it is not possible to acquire or analyze respiration signal but it is possible to obtain information on sleep macrostructure.