Research On The Complexity Of Schizophrenia’s Sensory Gating Based On Fuzzy Entropy

Schizophrenia affects the quality of life of human beings and even life safety seriously.Accurate diagnosis is the primary task of controlling and treating the disease.Due to the outstanding advantages of EEG signals,Traditional diagnostic methods mainly detect diseases from the time domain and frequency domain of EEG signals.However,these methods are not only time-consuming and labor-intensive,but also subjective,unable to effectively mine the more valuable hidden information,and there is always no uniform standard for the diagnosis of schizophrenia.Therefore,the research on the accurate diagnosis of schizophrenia has become particularly important.With the development of nonlinear dynamics and the nonlinear characteristics of EEG signals,the use of nonlinear indicators to analyze changes of complexity of brain disease has become a hot topic in the clinic.Previous studies have demonstrated significant complexity abnormalities in patients with schizophrenia.As a kind of non-linear index,entropy has strong anti-noise ability and small dependence on data length,and is widely used in disease diagnosis.While schizophrenia patients have significant sensory gating(SG)defects,SG-P50 has been shown to be an important internal phenotype of schizophrenia,and many researchers have used P50 evoked potentials to study sensory gating mechanisms.However,the change in the complexity of schizophrenic patients during sensory gating tasks remain unknown.Therefore,this study introduces entropy into the study of SG complexity,expecting to find kinetic evidence of abnormal SG in patients,and provide a reliable and accurate evaluation index for clinical diagnosis of schizophrenia.In this paper,the EEG data of SG were recorded in 55 normal subjects and 61 schizophrenia patients from the cooperative unit Beijing Huilongguan Hospital by using the auditory paired stimulation paradigm.After pretreatment,the approximate entropy,sample entropy and fuzzy entropy were used to investigate the changes in the baseline activity(BaseLine,BL)and the complexity of brain activity under paired stimulation(S1 and S2).The SG abnormalities in patients with schizophrenia were verified from the normalization and inhibition ratios.Finally,the "EEG image" is constructed by combining various characteristics of the EEG signal,and the deep learning method is used for classification research to verify the detection ability of the entropy.The findings provide a deeper understanding of patient sensory information processing and SG defects,providing scientific and effective testing standards for clinical diagnosis of schizophrenia.The main contents of this research are as follows:(1)Comparative study on approximate entropy,sample entropy and fuzzy entropy of their detection abilityThe approximate entropy,sample entropy and fuzzy entropy were used to study the complexity of the resting state and the task state in the normal control group and the schizophrenia patient group,and the detection ability of the three entropies was compared.Electrodes that are significantly different constitute the region of interest.(2)Research on SG anomaly mechanism based on fuzzy entropyExplore the study of the pattern of complexity changes in SG patients with schizophrenia.The differences between the two groups under paired stimulation were analyzed by normalization.The complexity suppression ratio is used to detect the difference between the two groups of SG,and the different electrodes are extracted to analyze the region of interest,and then the entropy of the region of interest is correlated with the patient’s scale information.(3)Classification of schizophrenia patients based on deep learningThe entropy value is used as the input characteristic of the classifier to further confirm the detection ability of the fuzzy entropy.Firstly,the detection performance of different entropy in traditional machine learning classification method is compared,and the detection performance of fuzzy entropy is the best.In order to further improve the classification performance of patients with schizophrenia,a deep learning algorithm combining time,space and frequency domain is proposed in the study.The fuzzy entropy value is used to construct the "EEG image",and the convolutional neural network(CNN)is trained to improve the classification accuracy.The above results consistently show that fuzzy entropy is an effective method to study SG complexity,provide kinetic evidence for finding patient’s SG defects,and can be used as a nonlinear standard for clinical diagnosis of schizophrenia.