Intelligent Diagnosis And Research Of Epileptic Disease Based On EEG

Epilepsy is a chronic disease due to abnormal discharge of neurons in the brain,that causes partial or whole brain dysfunction.EEG contains abundant information of brain function,which has a high reference value for the diagnosis of epilepsy.In the traditional diagnosis of epilepsy,it is usually necessary to collect patient's EEG data for a day or more,the large number of EEG data make the medical staff too tired,lead to reduce the detection efficiency,and there exist malpractice of subjective factors,the inspection in different standards.Therefore,the intelligent diagnosis of epilepsy is a particularly important research direction.At present,many methods of signal processing and analysis are used in the classification diagnosis of epileptic EEG,but there are problems of few categories and low classification accuracy.This dissertation combines with other literatures provide a classification method of epileptic EEG based on wavelet analysis,linear and nonlinear feature extraction,support vector machine and particle swarm optimization.The experiment result shows that the method used in this dissertation can effectively classify the EEG data into the health period,the intermittent period of epileptic seizures,and the epileptic seizures period,The specific contents of the dissertation as follows:First of all,I discussed the current situation at home and abroad of intelligent diagnosis of epilepsy,comparing the advantages and disadvantages of various research methods;the types of EEG and the characteristic waveform and frequency distribution of epileptic EEG.Secondly,the source of epileptic EEG data and the preprocessing of wavelet transform are introduced.After decomposing the original EEG with the 5 layer of wavelet,the EEG signal in the epileptic characteristic frequency band is obtained.Thirdly,considering EEG signal is chaotic signal which is an important conclusion,I extract the linear and nonlinear characteristics in the epileptic characteristic frequency band,including fluctuation coefficient,approximate entropy and sample entropy.As an important innovation in this dissertation,according to the energy distribution of EEG signals at different states and scales,adjusting the coefficient of each characteristic parameter,increasing the epileptic characteristic frequency coefficient with high energy ratio,and decreasing epileptic characteristic frequency coefficient with low energy ratio.After weighted each of characteristic parameters according to the energy coefficient,and formed the final eigenvectors.Finally,seven experiments are given in this dissertation.Support vector machine is used to classify and diagnose EEG signals.Different kernel functions are used,set up the feature vectors before and after parameters,The paramnet of support vector machine was optimized by using particle swarm optimization.This method has achieved better experimental results,and the accuracy rate of correct classification can reach 99.83%.The experimental results in this dissertation were compared with other literatures,it not only divides the EEG signals into three categories,but also improves the accuracy of classification,and it provides a new way of thinking for subsequent research.At the end of the dissertation,the user interface of the intelligent diagnosis system is designed.The main purpose of the interface is to simplify operation steps,to observe experimental simulation results and make it widely applied in practical operation.