Sleep apnea classification using ECG-signal wavelet-PCA features.

Sleep apnea is often diagnosed using an overnight sleep test called a polysomnography (PSG). Unfortunately, though it is the gold standard of sleep disorder diagnosis, a PSG is time consuming, inconvenient, and expensive. Many researchers have tried to ameliorate this problem by developing other reliable methods, such as using electrocardiography (ECG) as an observed signal source. Respiratory rate interval, ECG-derived respiration, and heart rate variability analysis have been studied recently as a means of detecting apnea events using ECG during normal sleep, but these methods have performance weaknesses. Thus, the aim of this study is to classify the subject into normal- or apnea-subject based on their single-channel ECG measurement in regular sleep. In this proposed study, ECG is decomposed into five levels using wavelet decomposition for the initial processing to determine the detail coefficients (D3-D5) of the signal. Approximately 15 features were extracted from every minute of ECG. Principal component analysis and a support vector machine are used for feature dimension reduction and classification, respectively. According to classification that been done from a data set consisting of thirty-five patients, the proposed minute-to-minute classifier specificity, sensitivity, and subject-based classification accuracy are 95.20%, 92.65%, and 94.3%, respectively. Furthermore, the proposed system can be used as a basis for future development of sleep apnea screening tools.