Automatic Derivation of Nocturnal Desaturation Burden from Oxyhemoglobin Saturation Signal

Sleep apnea is a common respiratory disorder which causes frequent reduction in oxygen level (hypoxemia) during sleep. Sleep apnea is highly undiagnosed due to the challenges associated with overnight polysomnography. Moreover, the measure of sleep apnea severity, called apnea-hypopnea index, is a poor measure in capturing the differences in sleep apnea related outcomes among different individuals. This is particularly important for surgical patients as sleep apnea is highly associated with postoperative respiratory depression and hypoxemia, especially during sleep. Recently, novel measures such as hypoxic burden were introduced to better quantify the burden of sleep apnea. However, calculation of hypoxic burden relies on manual annotations of respiratory events using polysomnography recordings. In this research, we present a technique which automatically detects desaturation episodes from oxyhemoglobin saturation signal and accordingly computes the burden of each desaturation episode as the area under its curve. Subsequently, we extracted features to characterize overnight desaturation burdens and investigated the association of the extracted features with sleep apnea assessment measures (apnea-hypopnea index, respiratory-related and total arousal index) and postoperative overnight respiratory depression and hypoxemia.