Automatic Sleep Apnea Syndrome Detection and Classification of Severity Level from ECG and SpO2 Signals

Abstract

Sleep apnea-hypopnea syndrome (SAHS) is widespread sleep and respiratory disorder which is characterized by breaks in breathing or instances of superficial or uncommon breathing during sleep. SAHS diagnosis is commonly performed using Polysomnography (PSG). However, this technique is a very complex and time-consuming procedure due to the need of many physiological variables and the use of multiple sensors attached to the patients the whole night. Moreover, PSG is also inconvenient as an expert human observer is required to work overnight and relies on a doctor‟s experience. Thus, the possibility of occurrence of the white coat effect, as children take longer to adapt to the hospital environment and to fall asleep, a fact that will affect the results. In order to improve the diagnosis efficiency, reduce the complexity and diagnosis time and ensure a more accurate diagnosis, a quantitative and objective method is required. Different biosignal features have been proposed for the detection of sleep apnea in the literature. However, in most cases, only a single biosignal system alone is used with the only aim of detection of sleep apnea. Using one signal alone may cause false positive or false negative results due to different artifacts. Using two or more signals simultaneously increases the reliability of the result. Sleep apnea syndrome detection based on a simultaneously recorded electrocardiograph (ECG) and saturation of oxygen (SpO2) signals, individually and in combination, has been proposed in this thesis. In addition, the automatic classification of sleep apnea severity level has been incorporated to enhance the diagnosis and treatment procedure. Various features from the RR intervals of ECG, and a number of statistical features from the SpO2, were extracted as indicators of sleep apnea. The features were then fed to support vector machine (SVM) for classification. An accuracy of 99.1%, specificity of 98.08% and sensitivity of 100 % has been achieved using the simultaneously recorded combination of two biosignal features and found to be better compared to other proposed techniques. Using the combined features is inherently more robust, as in the event of either channel being poor quality, the system can continue to make an analysis based on the other channel and achieve better accuracy compared to using either signal alone.