Characterization of EEG under anesthesia and a comparative study with BIS

This paper investigates the problem of automatic depth of sedation (DOS) estimation from electroencephalogram (EEG) recordings. Bispectral index (BIS) is used as a criterion to estimate the new modeling and calculation methods in this paper. Estimation of DOS contains two steps. First, time-frequency domain signal processing technique and nonlinear dynamical techniques based on the concept of chaos have been used to extract multiple features; Second, these features have been employed and modeled by Lasso technique, DOS is calculated and the evaluation models for specific EEG samples are obtained as a result. In emulation and clinical application, DOS are calculated and compared with BIS. Result shows that estimation including the nonlinear features are much closer to BIS than that without nonlinear features, it shows that the nonlinear chaos features have a great advantage in reflecting the chaotic characteristics of the brain during the anesthesia process.