A statistical complexity measure can differentiate Go/NoGo trials during a visual-motor task using human electroencephalogram data

TL;DR

This study demonstrates that statistical complexity measures derived from EEG data can effectively distinguish between Go and NoGo trials in a visual-motor task, revealing differences at both individual and group levels.

Contribution

The paper introduces a novel application of information-theoretical quantifiers, including entropy and complexity, to differentiate trial types in human EEG data using the Bandt-Pompe method.

Findings

Certain EEG channels reliably differentiate trial types.

Go and NoGo trials occupy distinct regions in the complexity-entropy plane.

Time windows with maximal differentiation are identified.

Abstract

Complexity is a ubiquitous concept in contemporary science and everyday life. A complex dynamical system is usually characterized by a blend of order and disorder, as well as emergent phenomena that often span multiple temporal and spatial scales. The information processes related to different cognitive processes in the brain can be studied in light of statistical differences based on complexity measures of the electrophysiological time series from different trial types. Recently, it has been demonstrated that a symbolic information approach can be a valuable tool for discriminating response-related differences between Go and NoGo trials using the local field potential of brain regions in monkeys. The method shows significant differences between trial types earlier than the simple average of the electrical signals. Here, we analyze human electroencephalogram data during a Go/NoGo task using information-theoretical quantifiers, including entropy and complexity. We employ the Bandt-Pompe symbolization methodology to determine a probability distribution function for each time series and then to calculate information theory indices. We show that a few channels, especially a central and an occipital electrode, consistently differentiate Go/NoGo trials at the individual level. We also show that different trial types occupy separate regions in the complexity-entropy plane. Our method also captures specific time windows to better differentiate the trial type. Moreover, we show that our results are robust at the group level, considering the average activity of all subjects.