Based On The Nonlinear Dynamic Characteristic Indexes Of Epileptic EEG Signals Classification Positioning And The Forecast Method Research

Objective: Epilepsy is a chronic neurological disease caused by abnormal electrical discharge of neurons in the brain,its sudden,transient,and repetitive nature causes great harm to the patient’s mind and body.Electroencephalogram(EEG)plays an important role in the diagnosis of epilepsy,type of epileptic seizures,and qualitative determination.It not only can be classified in the clinician to seizures,lesion location of the region to provide powerful basis,but also can guide the curative effect of epileptic drugs make objective evaluation.Therefore,in patients with epilepsy brain electrical signal analysis and processing,to explore the pathogenesis,and for the automatic detection,accurate diagnosis and treatment to provide very useful information.Methods: The platform is based on high-performance parallel computing,using Matlab software,and using correlation analysis to select the guides.The epileptic brain electrical signals were decomposed by the method of empirical mode decomposition,By using the method of principal component analysis of Hurst index,Lyapunov index and sample entropy,permutation entropy and wavelet entropy nonlinear dynamic index of five dimension reduction,and extract the principal component in comprehensive consideration,by using single index and integrated index respectively on the electrical components and characteristics of decomposition,using two classifications for brain electrical signals(interphase and episode)were performed using the cost sensitive support vector machine and random forest.Using statistical software to analyze the characteristics and classification results;Using the medical reference range method,we tried to predict the location of the epileptogenic foci and the time of onset.Results: 1.Using correlation analysis to select the number of guides,it is possible to minimize the correlation between the various guides and the characteristics of brain neuron activity.2.The main component analysis can effectively extract the characteristics of the electrical signals in the brain byreducing and synthesizing the eigenvectors.3.The combination of nonlinear dynamics,machine learning and statistics has established a model of identification,positioning and prediction of epileptic brain electrical signals.4.In the recognition of electrical signals in the brain,the classification accuracy of the former is higher than that of the single feature index.5.Both the cost sensitive support vector machine and the random forest can effectively classify the epileptic brain electrical signals in two classifications,and there is not much difference between the two classifiers.6.On the basis of high performance parallel computing and nonlinear dynamics and medical statistics can be quickly and efficiently and realize the combination of epileptogenic zone epilepsy and seizure prediction.Conclusion: This study is based on high performance parallel computing,the correlation analysis of guide number selection,principal component analysis and the multiple nonlinear dynamic indexes,using machine learning method of epileptic eeg signals effectively,which can identify epileptogenic zone combined with statistical implements for the initial positioning and attack the prediction of time,for clinical provides a new method for automatic recognition of eeg signals,for the treatment of patients with epilepsy have provided the certain theory basis,has a certain practical value and social significance.