Oxygen Saturation Sample Entropy, a Novel Diagnostic Tool in Sleep Disordered Breathing

Abstract Purpose To assess whether a Network Physiology approach using Sample Entropy and Transfer Entropy can be applied to simple physiological signals obtained during sleep studies, to accurately distinguish between different types of sleep disordered breathing (SDB) and streamline the diagnostic process. Methods Retrospective study on patients who underwent a sleep study between January 2017 and December 2021. The training dataset, used for algorithm development, included four clinically important groups: normal, obstructive sleep apnoea alone, sustained nocturnal hypoxemia with a high AHI (≥ 30/hr) and sustained nocturnal hypoxemia with a low AHI (< 30/hr). Mean, standard deviation, Sample Entropy and Transfer Entropy was calculated for heart rate, respiratory rate, oxygen saturation and nasal flow for each patient. Sample entropy is a measure of signal complexity. This was validated in a separate test dataset. ROC analysis was used. Results In the training dataset (n = 105), the Sample Entropy of the oxygen saturation signal was significantly different in patients with SDB compared to normal studies. The area under a ROC curve for predicting obstructive sleep apnoea from normal studies and sustained hypoxia with a high AHI (≥ 30events/hr) compared to sustained hypoxia with a low AHI (AHI < 30events/hr) was 0.943 and 0.785 respectively. When tested in the test dataset (n = 80), oxygen saturation Sample Entropy above 0.1456 was 100% sensitive and 60% specific in diagnosing obstructive sleep apnoea. Patients with OSA had significantly increased Transfer Entropy from HR → SpO2, RR → SpO2 and NF → SpO2; and significantly decreased Transfer Entropy from SpO2 → RR. Interpretation Network Physiology mapping of oxygen saturation can help distinguish between different types of sleep disordered breathing and has the potential to support simplified global diagnostic pathways for sleep apnoea utilising oximetry alone.