Research On The Eeg Entropy Algorithm And Anesthesia Monitoring Application

People have been studying the effect of anesthetic on the brain since1940s.However, assessing the depth of anesthesia (DoA) and determining the relationship ofanesthesia and consciousness have always been issues to researchers and anesthetists.Most commercially available anesthesia monitor products lay on linear theories, and havemany limitations in extracting EEG characteristics. It has been reported that thesemonitors exist problems such as false alarm, time delay and fuzzy values. Besides, due tothe high expense of these monitors, most hospitals in our country could not afford thecost of equipping them. Therefore, to develop a DoA monitor of our own intellectualproperty is not only theoretical, but also practically significant. In this paper, the authorperformed in-depth analysis of the EEG signal in anesthesia, compared the pluses andminuses of current available DoA monitors, and proposed a DoA monitoring methodbased on EEG entropy. Based on this method and other parameters, a DoA monitor wasdesigned.Firstly, in view of current researches and documents, the mechanism of anesthetic'seffect on the brain was discussed. The benefits and drawbacks of existing DoA monitorswere discussed. The response of the brain to different anesthetic agents was and themechanism of conscious generation and action was analyzed. The design principles ofDOA monitors, the considerations of quantization of DoA indices, algorithms,optimization approaches have been discussed in detail. The algorithm complexity,response time, artifact removal have been compared elaborately.Secondly, three non-linear algorithms of assessing DoA were purposed, namelypermutation entropy, Hurst index, wavelet sparsity. Among these methods, thepermutation entropy after outlier value removal by the general extreme studentizeddeviate (GESD) had the highest correlation with BIS index. After optimization by themaximal overlap discrete wavelet transform (MODWT), the Hurst index also had goodresults. The pharmacokinetic/pharmacodynamics (PK/PD) modeling and the predictionprobability showed that all these methods could predict the effective anesthetic drug concentrations. Considering the computation complexity, permutation entropy wasadopted as the main parameter of assessing DoA.Then, for conscious status evaluating, a bi-channel information coupling strengthmethod was purposed. Analysis showed that the mutual information based onpermutation patterns was more persistent to artifacts and had a better ability in describinginformation coupling. As anesthesia deepened, the permutation-based mutual informationdecreased. This confirmed the hypothesis that information coupling between differentbrain areas weakens when anesthesia deepens. Auto mutual information analysis basedon permutation also revealed that the self-coupling of EEG weakened when anesthesiadeepened. The information coupling analysis based on permutation mutual informationcould distinguish different conscious status, and is of great importance to anesthesiamonitoring.