Causal influence: Advances in neurosignal analysis

The analysis of multichannel recordings such as electroencephalography (EEG) and magnetoencephalography (MEG) is important both for basic brain research and for medical diagnosis and treatment. Multivariate linear regressive analysis such as the AutoRegressive (MAR) modeling is an effective means to characterize, with high spatial, temporal, and frequency resolution, functional relations within multichannel neuronal data. Recent advances in MAR modeling show promise for the analysis and visualization of large-scale network interactions, especially in the ability to assess their causal relations. This article provides a detailed review of the advances in the development and application of causal influence measures for analyzing neurosignal within the framework of the MAR spectral analysis. First, we outline mathematical formulations of the MAR model and its related estimation procedures, with emphasis on the development of causal influence measures for analyzing brain circuits. Second, we address the technical issues on the practical applications of the causal measures to the neurobiological data. Of particular interest is the recent development of adapting the MAR to analyze neural spike train data. Third, we present a variety of applications ranging from basic neuroscience research to clinical applications as well as functional neuroimaging. We finally conclude with a brief summary and discuss future research development in this field.