Detection of sleep apnea events using electroencephalogram signals

Abstract

Sleep apnea is breathing disorder that leads to other disorders related to the brain and heart. This paper proposes detection of sleep apnea using a single feature Lampel-Ziv complexity of electroencephalogram (EEG) signals. Firstly, tunable-Q wavelet transform (TQWT) analyzes EEG signal into sub-bands (SBs). The Lampel-Ziv complexity (LZC) feature is computed from each SB for the discrimination of sleep apnea and control events. The Kruskal–Wallis (KW) test is applied to assess the discriminative performance of LZC feature. The statistically significant LZC feature is applied to discriminant analysis, decision tree, and ensemble classifiers for the detection of apnea events. The ensemble classification technique subspace-K-nearest neighbor provided the best classification accuracy of 96% for apnea events identification. The other classification performance measures sensitivity, specificity, F1-score, and Matthew’s correlation coefficient are also attained higher values for the proposed method.