Recognition Of Brain Disorder On Nonlinear Hidden Information And Network Feature

Human brain is a highly sophisticated nonlinear system,using electroencephalogram(EEG)to explore the mechanisms of brain has gotten great progress and wide application in biomedical engineering and cognitive research.The activity of human brain displays a wide range of activation patterns during both normal and abnormal states,which are represented by the specific activity from the perspective of temporal,spatial and spatio-temporal domain.With the social development and influence of pressure as well as emotional crisis,the pathogen spectrum has changed a lot.The mental factors have become one of the major causes which influence our health greatly.Due to the dynamic changes of the brain could be reflected by the physiological signals from the perspective of spatial and temporal pattern,using EEG to diagnose and identify the abnormal state of neurological disease has become a hot topic in the relevant area.Traditional linear and nonlinear thechniques have played vital role in the research.There are still some limitations in the exploration of brain disorder,since they could not discern or characterize subtle changes of the dynamic behavior.By analyzing weak EEG signals,a strategy which utilizes nonlinear meaures to explore hidden information to diseren epilepsy disorder and network features from the perspective of spatial domain to discern brain disorder of conversion blindness,is adopted in this research,and further experiments are attempted to evaluate the proposed methods.This research is studied from the following four aspects.1.A wavelet-entropy method is proposed which used the wavelet coefficients to determine the scale factor of the measure of sample entropy.Based on the potential markers of abnormality discovered using the wavelet-Entropy method,a novel model is developed for epileptic seizure detection.2.A novel FRFT-chaos method is proposed.The traditional univariate chaos measure does not effectively identify multiple states of epileptic EEG.Taking account of the characteristics of epileptic EEG signal and the fractional Fourier transform(FRFT),which is sensitive to specific transients and could provide compact support in appropriate transform order,the model of preictal prediction by using FRFT-chaos is developed by combining FRFT and chaos meature.Algorithm of the largest Lyapunov exponent is modified to adapt the transformed time series in complex space,and by using an energy measure to determine appropriate fractional order,subtle chaotic dynamics of epileptic signals could be captured in FRFT domain.Experimantal results with scalp EEGs demonstrated the potential and robustness of the proposed method,and the performance indices are improved.3.A method which fuses coarse key network metrics with traditional network features is proposed.To complement the incompetence of global or local network feature in differentiating abnormal states of relevant diseases,coarse key characteristics are selected by algorithm of minimum redundancy maximum relevance.In applying it to research functional connectivity network of conversion blindness patients,performance elevation is achieved by the proposed approach,which provides new perspective to discern and understand brain disorder of conversion blindness.4.A composite approach that integrates amplitude and phase information on coherence is proposed,which might overcome the deficiencies of the single information source.In applying it to research functional connectivity network of conversion blindness patients,minimum spanning tree topology is helped to explore hidden network information,when fusion with traditional network feature it yields higher accuracy in differentiating conversion blindness from controls.Alterted network configuration of conversion blindness brain is discerned from different perspective by such a compsosite approach.Additionally,region of brain dysfunction associated with conversion blindness is located,which might help improve the clinical care and treatment of patients.