Análise de dados eletrofisiológicos e de simulação computacional utilizando a metodologia de simbolização de Bandt-Pompe

Abstract

The brain is a nonlinear dynamic system in which small variations in external stimuli can trigger very different responses. To better understand it, time series analysis is a powerful approach that, through statistical tools, allows us to extract relevant information about the dynamics of this system. This makes it possible to identify patterns associated with different cognitive and behavioral states. In this work, we use electrocorticography (EcoG) data, freely available on an open network, from a human patient exposed to two different visual tasks. We also analyze time series data obtained through simulations performed with the CROS (Critical Oscillations) computational model. For the analyses, we use the symbolic methodology of Bandt & Pompe, based on Causal Information Theory, and calculate the Probability Distribution Function (PDF) for different combinations of elements in a time series. To do this, we consider the dimension, the delay, and the direction of the analysis. With the obtained PDF, we calculate the quantifiers of Information Theory: Shannon Entropy H and Statistical Complexity C. From these measurements, it is possible to distinguish the two cognitive activities: periods of visual search, associated with an active task, and blank screen intervals, corresponding to waiting periods. In addition, for the series obtained by simulation, we observe how different values of the excitatory connectivity parameter (rE) influence the dynamics of the system. In both applications of the Bandt & Pompe method, we identify which time scales are most relevant for information processing, which is the basis for the cognitive functions of the brain.