A Computationally Efficient Method for Brain Information-theoretic Based Causality Detection Using Multichannel EEG

Information flow or causal interaction between neuronal populations of the brain is a critical issue in describing the dynamics of such a complex network, which can be best described by the illustrative features of graphical modeling. In this paper, we exploit the information-theoretic based causality detection measures to propose a uniform framework to derive a graphical model for the statistical analysis of multivariate processes from observed time series. Here, our main focus is on the efficient calculation of the measures for link estimation through searching for the most informative variables that drastically reduces the estimation dimension. We demonstrate the performance of our method for stationary processes using numerical simulations of nonlinear processes.