Is detection of different anesthetic levels related to nonlinearity of the electroencephalogram?

Nonlinear electroencephalographic (EEG) parameters for monitoring depth of anesthesia may be superior to methods based on the linear power spectrum of the signal, as they may reflect additional information related to the dynamics of the cerebral cortex. The present investigation evaluates, whether nonlinear parameters of regularity Correlation Dimension (D2) and Approximate Entropy (ApEn) detect characteristics beyond a linear structure of the EEG during anesthesia. For this purpose, the parameters were calculated on EEG signals and on surrogates of the EEG which preserve the linear structure together with underlying dynamics of a Gaussian linear stochastic process. The formal test may indicate occurrence of nonlinearity at specific anesthetic levels. Furthermore, the ability of ApEn and D2 to separate five different anesthetic levels from awake to deep anesthesia was compared to Spectral Entropy (SpEn) and spectral estimation of Hurst Exponent (HE) which are linear measures of regularity. Results based on EEG data of a volunteer study show, that ApEn and D2 detect nonlinearity and separate the anesthetic levels with higher prediction probability than SpEn and HE which are both not influenced by nonlinear information of the EEG.