Depth of Anesthesia Monitoring Method Based on EEG Microstate Analysis and Hidden Markov Model

Abstract

Electroencephalogram (EEG) microstate analysis is an important emerging method that can classify continuous multichannel EEG signals into a limited number of microstates through clustering. Microstate analysis combines the time and space information of EEG, which can reflect important transformation process of high-level cognitive functions in the brain. In recent years, Microstate analysis has made great progress in the research of depth of anesthesia (DOA) monitoring. In this paper, a new DOA monitoring algorithm is designed by combining microstate sequence and hidden Markov model (HMM). The trained Hidden Markov Model shows the information of brain nerve activity hidden in the microstate sequence, which can effectively distinguish the mental states of different DOAs, thereby realizing the corresponding DOA classification. The experimental dataset was obtained from an open-access section of the University of Cambridge Data Repository, which contains EEG data from 20 healthy subjects. During propofol injection, the brain states of the subjects were divided into four conditions: baseline (BS), mild sedation (ML), moderate sedation (MD), and the recovery stage (RC). The algorithm classified BS and ML, BS and MD, ML and MD with the accuracy rates of 71.40%, 73.48%, 67.75% respectively. This shows that the microstate analysis has great application potential in the study of anesthesia. Hidden Markov model training for microstate sequences can become a new research direction for DOA monitoring.