A sleep staging algorithm for SAHA detection based on C4/A1 EEG

SAHS event accurate detection is vital for sleep monitoring and auxiliary ventilation therapy. Studies have shown that 63.71% SAHS false-positive error occurred during sleep awakening period. This paper puts forward a kind of sleep state recognition algorithm based on single channel C4/A1 EEG to Complete the screening of monitoring data of oronasal flow and abdominal displacement, which can raise the accuracy of SAHS detection and classification. Considering the comfort of long-term sleep monitoring, this paper choose EEG signal, oronasal flow signal and abdominal displacement signal as the research object. The fifth-order Daubechies wavelet is used to extract C4/A1 single channel EEG rhythm signal. Then the 12-dimensional feature vector is constructed from the EEG rhythm energy ratio, sample entropy mean value and variance. After dimensionality reduction in Principal Component Analysis (PCA) analysis, sleep/wake classification is performed using SVM, so as to complete data screening of oronasal flow and abdominal displacement data. 3000 sample dataset of 30 second C4/A1 EEG data were randomly selected from 15 subjects. The proportion of sleep and wake is 1:1, and the proportion of training set and test set is 2:1. The results of sleep/wake classification shows that the average accuracy of training set is 89.0%, and the average precision of training set is 89.6%, and the average accuracy of test set is 88.6%, and the average precision of test set is 89.2%. The Support Vector Machine (SVM) model is used to effectively realize the sleep/wake classification for SAHS detection. The overall classification accuracy is 83.94%, the average sensitivity rate is 86.07%, and the false positive rate decreases from 28.19% to 5.34%, indicating a significant improvement. The proposed sleep staging algorithm can improve SAHS detection effectivity and which may provide technical support for subsequent realtime monitoring of SAHS and personalized assisted ventilation therapy.