EEG-based classification of consciousness during sedation using global spectra principal components

The objective of sedation is to maintain patient safety, and reduce the anxiety or pain during surgical procedure. In this point of view, method for monitoring the depth of anesthesia (DOA) should be reliable. Previous electroencephalogram (EEG) based DOA studies under general anesthesia (GA) have shown the significant correlation between brain state and neurophysiological characteristics. However, no matter how many existing DOA studies are under GA environment which is considered as 'the deepest sedation', it could not clearly distinguish between consciousness and unconsciousness during sedation. In this paper, we proposed a novel feature extraction technique, called global spectra principal component (GSPC) motivated by global field synchrony (GPS), using channel-wise coefficients from multi-dimensional channels in interest frequency ranges. As a result, average classification performance of 25 subjects represented 98.7±2.1%. It showed that the proposed method was an efficient feature extraction technique for classification of 'consciousness' and 'unconsciousness' even during sedation.