A Data-Driven Clinical Decision Support System for The Diagnosis of Sleep Apneas

Abstract

Sleep apnea syndrome (SAS) is a sleep-related respiratory disorder that has an important consequence on human being’s health, which is characterized by the occurrence of five of more apneic events (apnea or hypopnea) per hour of sleep. To enhance early detection, our paper proposes a prototype of a clinical decision support system (CDSS) for the diagnosis of sleep apnea syndrome (SAS), based on the automated analysis of the blood oxygen saturation, heart rate and respiratory signal instead of in-hospital overnight polysomnography (PSG). A web-based architecture system taking advantage of information technology was proposed, which consists of sleep analysis, respiratory analysis, and diagnosis analysis of three sub-systems. In sleep analysis, classifier of sleep and wake under logistic regression (LR) has been developed, with the dynamic time warping algorithm introduced in feature extraction. In comparison with the classifier based on neural network and hidden Markov model respectively, LR classifier can acquire a better performance. By means of respiratory analysis, sleep apnea, and hypopnea event can be recognized and classified based on the analysis of respiratory signals and blood oxygen saturation. Combing the results from sleep analysis and respiratory analysis sub-system, Apnea Hypopnea Index (AHI), diagnosis basis of SAS, can be calculated in the diagnosis analysis sub-system. According to the AHI and the physical features of patients, such as BMI, height, weight, and age, patients’ condition can be assessed and some continual advice about whether the patient should do a detailed inspection on a PSG equipment will be given. The proposed system can provide essential and complementary information to assist in SAS diagnosis and help the clinician make proper decisions. Additionally, it can be used for home-based monitoring of suspected apneic subjects for a valuable early discovery.