Classification of midazolam-induced sedation depth based on spatial and spectral analysis

Distinction of loss and recovery of consciousness is an important component in consciousness study. To find transitions in and out unconsciousness, monitoring depth of anesthesia (DOA) should be reliably assessed. Previous studies have proposed several methods for measuring DOA, and one of the significant methods to identify between awaked and anesthetized state is global filed synchrony (GFS). GFS used the coherence information from the global electroencephalogram (EEG) channels by using the effects of phase and amplitude relationship simultaneously. However, most recent work showed that there were specific independent EEG amplitude as a biomarker of consciousness while changing the transition into and out unconsciousness. In this paper, we proposed a GFS based feature extraction technique, using coefficients of multi-dimensional channels in interest frequency range in repeated sedation condition. It allows to extract significant spatial and spectral features. We classified the 'wakefulness' and 'unconsciousness' from midazolam-induced sedation and linear discriminant analysis (LDA). As a result, classification performance in 25 subjects represented 97.09%. Also, it showed that the proposed method was an efficient feature extraction method for classification of 'wakefulness' and 'unconsciousness'.