Sleep Stage Classification Using Variational Mode Decomposition and Wrapper-Based Feature Selection From the Single Channel EEG

Sleep stage classification can diagnose various sleep disorders and sleep patterns. The classification model classifies many stages of sleep, including wakefulness, non-rapid eye movement (NREM) sleep, and rapid eye movement (REM) sleep. Precise and reliable sleep stage classification is crucial for clinical applications and research studies. Diagnosing sleep disorders without an accurate automated classification model is laborious and susceptible to inaccuracies.This may lead to delayed or ineffective treatments. Manual scoring is tiresome and inconsistent, making it difficult to provide personalized treatments to treat sleep diseases efficiently. Sleep disorders, including narcolepsy, insomnia, and sleep apnea, can be identified and monitored by automatic sleep classification. The proposed framework uses variational mode decomposition (VMD). The electroencephalogram (EEG) is processed into band-limited intrinsic mode functions (IMFs) by VMD. Each IMF signal in EEG was broken down into 15 features based on time, frequency, and information theory. Furthermore, the optimum feature subset was selected using the Wrapper-Based Feature Selector (WBFS). Finally, well-known classifiers used to classify the EEG signal into five distinct sleep stages. This study achieves accuracies of 94.84% and 95.20%on the Sleep-EDF database, and 95.60% and 96.17% on the ISRUC-Sleep dataset, for the balanced and unbalanced cases, respectively.