Quantifying Evoked Responses through Information-Theoretical Measures

Abstract

Being the most sophisticated information processing apparatus known to us, understanding the brain offers great possibilities. A viable candidate to advance the understanding of this cortical information processing machine is information theory. With its universal applicability, it enables the modeling of complex systems, is free of any requirements about the data structure and can infer the present underlying brain mechanisms. Algorithmic information theory seems particularly suitable since it can estimate the information contained in individual brain responses. Here, we propose a measure grounded in algorithmic information theory termed Encoded Information as a novel approach to analyze neurophysiological recordings. Specifically, it enables an assessment of the encoded information mean responses share with one another by compressing the respective signals. By applying the approach to five cortical activity types originating from intracranial electroencephalography recordings of humans and marmoset monkeys, we demonstrate that the information-based encoding can compete with conventional approaches such as the t-test or mutual information. Information-based encoding is attractive whenever one is interested in detecting where in the brain the neural responses differ across experimental conditions.