Study On Analysis Of Neural Oscillations During General Anesthesia

The key issue in surgery is to determine the appropriate depth of anesthesia (DoA)during surgery. As yet, although the mechanism of general anesthesia remains largelyunclear, many observations revealed that anesthetic medications all alter the synapticfunction, which produce the electroencephalographic (EEG) signals in the cerebral cortexor on the scalp surface. As EEG contains a wealth of neural oscillations with differentfrequencies, analysis of these neural oscillations may reveal the drug effects on the entirecentral nervous system, which can be viewed as a basis for developing the DoAmonitoring system and further understanding and exploring the mechanism of anesthesia.This thesis was devoted to the analysis of neural oscillations during anesthesia, and thefollowing work were conducted.Firstly, based on the Fourier spectral entropy, a novel measure called theHilbert-Huang spectral entropy (HHSE), which exploits the empirical mode decompostion(EMD), was proposed to study the complexity in frequency domain of EEG duringanesthesia. HHSE was applied to the EEG data obtained from patients in sevofluraneanesthesia, and its superiority was verified through the pharmacokinetic-pharmacodynamic (PKPD) modeling and prediction probability, etc. Further, as theensemble EMD (EEMD) can account for the mode mixing in the original EMD, theEEMD, as a substitute for the EMD, was exploited to calculate the HHSE. Unfortunately,since the EEMD is highly compute-intensive, a parallelized EEMD method was developedusing the graphics processing unit (GPU). Results demonstrated that the EEMD-basedHHSE performed better than the EMD-based one, and the GPU dramatically improved therun-time performance compared to the serial implementation of EEMD.Secondly, based on the permutation entropy (PE), a novel measure combiningmultiscale PE information called the CMSPE was proposed to quantify the dynamicalcharacteristics inherent in the anesthetic EEG on multiple scales. Simulated EEG seriesduring awake, light and deep anesthesia were used to select the parameters for themultiscale PE analysis: embedding dimension, lag time, data length and scales to be integrated into the CMSPE index. The CMSPE index was applied to EEG recordings frompatients who received sevoflurane anesthesia, and it outperformed the raw single-scale PEthrough the PKPD modeling and prediction probability analysis, etc.Thirdly, based on the general harmonic wavelet transform, an improved waveletbicoherence (WBIC) was developed using a phase randomization and surrogate dataanalysis, to study the cross-frequency coupling in EEG signals during anesthesia. Ninepotential EEG indices were obtained from the wavelet bicoherence matrix to quantify theeffect of isoflurane on the EEG. Results demonstrated that the WBIC-based indicestracked anesthetic effects better than the Fourier bicoherence-based ones, and the optimalWBIC-based index was determined through multiple quantitative performance. Further,the relationship between the patterns of wavelet bicoherence and the isoflurane end-tidalconcentrations was examined in detail. It was found that isoflurane caused thecross-frequency coupling between oscillation and slow oscillation, and increasingisoflurane concentration slowed the frequencies where the coupling occurred. Thisphenomenon of-coupling suggests that slow cortical oscillations organize the higheroscillation, which is consistent with other studies in natural sleep.Finally, a global synchronization index, which utilizes the S-estimator and coherenceanalysis, was proposed to explore the synchronization at multiple frequency scales amongmultiple channels of electrocorticographic (ECoG) signals in sheep. Through the statisticalanalysis of patterns of synchronization during desflurane, sevoflurane, isoflurane andenflurane anesthesia, it was found that the global synchrnization increased with theinduction of these anesthetic drugs, especially at the and β frequency bands, and thefrequency band where the synchronization changed significantly most was related to thespecific anesthetic drug used. This phenomenon of increasing synchronization may beexplained in terms of the decreasing of cortical information capacity.