Nonlinear Analysis Of Neural Oscillations During Anesthesia

Neural oscillations play an important role in higher functional activity of the brain.These oscillations are mechanically generated by the individual neuron or the interactionbetween neurons, so the oscillations are everywhere in brain from the micro to the macrolayer. The synchronous activity of numerous neurons can be described byelectroencephalography (EEG) in the macro scale. The tremendous number of neurons andthe complex network the neurons compose make the EEG signals strong nonlinear andnonstationary that the conventional linear time and frequency methods can not deeplydepict. Therefore, in this article, the nonlinear methods including entropy indexes andfractal methods were applied to analyze EEG signals during aneathesia. Thepharmacokinetics and pharmacodynamics (PK/PD) model and prediction probability (Pk)were used to assess the performance of the indexes.Firstly, as for the artifacts in EEG signals, the FIR filter was used to remove lowfrequency such as baseline drift and high frequency including electromyogrphy (EMG)noise. As for entropy algorithms, the sample entropy(SampEn), fuzzy entropy (FuzzyEn)and wavelet entropy (WE) based on wavelet transform were first applied to estimate thedepth of anesthesia (DOA). Besides them，eight entropy algorithm including shannonentropy (ShEn), spectral entropy (SpEn), permutation entropy (PE) and Hilbert–Huangspectral entropy (HHSE) were compared in monitoring the DOA. The PK/PD model, Pkand correlation analysis with BIS were adopted to assesse the entropy indexes. Mostresults showed the PE is better than other entropy indexes in aspects of noise immunity,tracking ability and calculation speed.Secondly, based on the fractal theory, the detrended fluctuation analysis (DFA) wasemployed to analyze the anesthesia EEG signals. The DOA is estimated by the changes offractal index. Futhermore, the multifractal detrended fluctuation analysis (MFDFA) waspresented to the EEG signals. The results showed the EEG signal during anthesthesia hasmultifractal characteristics, and the deeper the DOA, the stronger the intensity ofmultifractal. Finally, all the indexes metioned in the article were used to detect the burstsuppression pattern(BSP), which may be appeared in deep anesthesia. It was found thatonly ApEn and SampEn can detect the presence of BSP.