Research On Method Of Depth Of Anesthesia Based On Electroencephalogram

Anesthesia plays an important role in modern surgery,but improper use of anesthetic drugs can result anesthetic state too deep or too shallow.Whether anesthesia state is too deep or too shallow can cause great harm to the patient’s health.If anesthesia is too shallow,it will cause the patient to wake up during the operation and cause harm to body and mind of patient.If the anesthesia is too deep,it may cause delayed recovery,impaired physical function,and even threaten life of patient.Therefore,accurate and non-invasive monitoring depth of anesthesia(DOA)can not only guide the anesthesiologist’s medication,but also ensure the safety and effectiveness of the operation to provide a comfortable operating environment for patients.However,there is no uniform standard for monitoring DOA.Brain is affected by anesthetic drugs directly,so use electroencephalogram(EEG)to monitor DOA has attracted the attention of researchers.Based on EEG signals data of patients,a new method of monitoring DOA was proposed in this paper.Based on the EEG signals of 30 patients with general anesthesia,anesthesia state was divided into awake,shallow anesthesia,and anesthesia maintenance.The anesthesia maintenance is further divided into BIS value greater than 40 and BIS value less than 40 according to BIS values..Firstly,signal to noise ratio(SNR)was improved by preprocessing EEG signals.Then,the characteristic values related to DOA in EEG signals were extracted.Finally,according to the time of each anesthesia state during anesthesia surgery is inconsistent,the weighted k-nearest neighbor(WKNN)algorithm was selected.A complete monitoring DOA algorithm system is formed.The specific work content is as follows:(1)Aiming at the physiological signals of the human body,the power frequency interference of the electrical equipment and the baseline drift,a filtering algorithm based on wavelet decomposition,low-pass filter and empirical mode decomposition(EMD)isused.In this way,the interference contained in the EEG signal is filtered,and the SNR of the signal is improved.(2)Extracting useful feature vectors from EEG signals based on Detrended fuctuation analysis(DFA),Autoregressive model(AR model)and Sample Entropy(SampEn).Then,the distribution of these three feature vectors at different anesthesia depths was compared,and the correlation between three feature vectors and the depth of anesthesia was analyzed.(3)According to the uneven distribution of sample points,the WKNN classifier is selected.The optimal K value was selected by optimization and three types of feature vectors were used to classify DOA.DOA is divided into awake,shallow anesthesia,and anesthesia maintenance.The anesthesia maintenance is further divided into BIS value greater than 40 and BIS value less than 40 according to BIS values.Finally,the depth of anesthesia is divided into four categories and the classification rate is calculated.(4)By analyzing the results of classification of DOA by three types of feature vectors,AR coefficient and sample entropy were selected to classify the DOA.Through the optimization AR coefficient,2~5-order AR coefficient and sample entropy were finally selected.Then,the classification of DOA was done using the WKNN classifier.The final analysis results and data show that the correlation coefficient between SampEn of EEG signals and DOA is more than 0.8,which has a strong correlation.There is a correlation between the second-order to fifth-order AR coefficients and DOA,and the highest correlation coefficient is about 0.8.According to the classification results of WKNN,the classification accuracy of this method is as high as 88.71%.The method proposed in this paper can well complete the monitoring of DOA.