Biomedical Signals Analysis Based On Nonlinear Granger Causality

Biomedical signal analysis has been occupying a high position in medical research, and the rapid advance of computer and digital signal processing technology provides a technical feasibility for biomedical signal analysis and application. The development of digital signal processing technology is often closely related to the development of medical signals. Both of them can promote the development of biomedical signals in feature extraction, detection and clinical application. The development of biomedical signals also allows us to recognize and understand our own body. The causal relationship between the biological signals can reflect the link in the physical network. In order to reduce the complexity of the physiological network, this paper uses nonlinear Granger causal algorithm to calculate the causal relationship between physiological signals, so as to infer the causal effect between physiological network. The contents of the work are summarized as follows:First,This paper proposed nonlinear Granger causality method based on kernel method, which extending Granger causality to nonlinear space. In this paper the method is used to analyze the causal relationship between epileptic EEG and ECG signals. At present, the mainstream method of signal causal analysis is proposed by Granger, but in terms of physiological signal nonlinear, Grainger causal theory is no longer applicable. In order to promote the further research on the relationship between physiological signals. In this paper, kernel method for pattern analysis is used to extend Granger causality to nonlinear space, which is used to analyze the causal relationship between epileptic EEG and ECG. By comparing the coefficient of causality between EEG and ECG,it shows that EEG has significant influence on ECG, compared with the calculation results of normal samples, the influence in patients with epilepsy is weak.Two, This paper analyze the causality of sleep physiology signals based on the nonlinear Granger causality. This paper selected EEG, ECG and blood pressure signals during the period of sleep and awake. The nonlinear Granger algorithm is used to analyze the causal relationship between the three kinds of signal. Through the comparison of causal index in sleep and awake period, it shows that EEG has significant influence on ECG, EEG has significant influence on blood pressure and ECG has significant influence on blood pressure. The influence during sleep stage is more obvious, which shows physiological signals during sleep period reflect the physiological characteristics more accurately.Third, In this paper uses fuzzy kernel function to realize nonlinear Granger causality method, which is use to analyze the causal influence between EEG and ECG of samples. Gaussian kernel, sigmoid kernel function and the corresponding fuzzy kernel are used in the nonlinear Granger causality algorithm to analyze epileptic EEG and ECG. The results of the study show that using four kinds of kernel functions can find significant distinction of causal influence between EEG and ECG. Compared with the ordinary kernel function, using fuzzy kernel can save simulation time significantly. In addition, because the sigmoid function is a conditional kernel function, fuzzy Gaussian kernel function has better universality.