Automatic Diagnosis of Sleep Apnea and Its Type Using Linear Discriminant Analysis and Support Vector Machine Based on the Characteristics of Two ECG and PPG Signals

The main purpose and motivation of this research was to design and present a system for Automatic diagnosis of sleep apnea and its type. According to the characteristics of electrocardiogram (ECG) and photoplethysmogram (PPG) signals and how this disease is related to different characteristics in these signals, sleep apnea and its type can be recognized automatically with appropriate processing. In this research, PPG signals were first recorded, followed by pre-processing and noise removal on both ECG and PPG signals. Mean, median, mode, maximum, minimum, range of changes, standard deviation, variance and absolute deviation of the mean and absolute deviation of the mean for the linear discriminator and also extracted from 5 features of the HRV signal, which include the average interval of RRs, SDNN, VLF (frequency < 0.04 HZ), LF(0.04 < freq < 0.15) and HF (0.15 < freq < 0.4) are used for support vector machine. Then, some time characteristics were extracted from these three signals and some frequency characteristics were extracted from the Heart Rate Variability (HRV) signal. Then the optimal features were extracted by analysis of variance (ANOVA) method and two classifiers linear discriminant analysis (LDA) and Support vector machine (SVM) were used. According to the results, it has been found that among these two classification methods, SVM method has higher discrimination accuracy than LDA. These two classifiers have been used in three classes, i.e. healthy, obstructive apnea and central apnea. The accuracy of the results using features selected through ANOVA was 67.2% for LDA and 76.7% for SVM.