Research On Nonlinear Algorithm Of Brain Effective Connectivity In EEG Signals

One of the major research contents of China Brain Project is brain science research oriented by exploring brain secret and conquering brain disease.It is one of the hotspots in the research field to study the brain work mechanism,and early diagnosis and preoperative evaluation of brain diseases through the brain connectivity algorithm.Epilepsy is the second most common disease in neurology.Its repeated seizures and sudden onset seriously affect patients’ life.About 30% of all patients suffer from refractory epilepsy,who can’t control and cure seizures with drugs.For these patients,the treatment is removing the epileptogenic zone by surgical procedures.Generally,applying the algorithms of brain effective connectivity to analyzing the patients’ EEG signals,can provide significant help for the accurate positioning of the epileptogenic zone of the preoperative assessment.Therefore,this paper focuses on the brain effective connectivity algorithm in EEG signals.Wiener-Granger Causality Index is a method of researching brain effective connectivity.It can test the Wiener-Granger causal relation between signals in time domain.It also extends to frequency domain and develops many widely used algorithms,such as DTF(Directed Transfer Function)and PDC(Partial Directed Coherence).However,brain is a very complex structure.DTF and PDC are based on AutoRegressive eXogenous(ARX)model,and can’t test nonlinear causal relation between EEG signals.In terms of this issue,this paper adopts the 2-D Nonlinear PDC(NPDC)and 2-D Nonlinear DTF(NDTF)based on 2-D Nonlinear ARX(NARX)model.They can test linear and nonlinear causal relations between signals in frequency domain.However,when these algorithms are applied to analyzing multidimensional signals,they can’t disthingush direct and indirect causal relations between signals.Therefore,in this paper,based on single input multiple outputs NARX(SIMO NARX)model,2-D NPDC can extends to 3-D NPDC,which can test linear and nonlinear causal relations and distinguish the direct and indirect causal relation.In the experimental part,firstly,we use an autoregressive model including linear and nonlinear causal relations to verify that 2-D NPDC and 2-D NDTF are capable of testing linear and nonlinear causal relations.Afterwards,we apply 3-D NPDC on simulation model and physiological model.The experimental results show that 3-D NPDC can test and distinguish the direct and indirect causal relations between signals robustly.At last,we apply 3-D NPDC to processing and analyzing the causal relation between EEG signals.