Study on neural variability change of depression and feedback processing by using permutation entropy

Abstract

Neural variability is a quantification of changes in neural activity, and it is known to reflect the random activities of neurons or nonlinear properties of neural networks. Therefore, various methods of measuring neuronal variability are being developed to understand the process of cognitive information processing or neuropsychiatric disorders. However, studies on neural variability changes have limitations of signal variability measures, such as Lempel–Ziv complexity. The recently developed refined composite multiscale permutation entropy (RCMPE) overcomes the limitations of existing signal variability measures, such as length dependence, and helps measure variability at multiple scales. The study was conducted using this RCMPE under the assumption that there will be a specific layer of neuronal variability that reflects the cognitive information processing or disease state. To confirm this hypothesis, in our first study, we identified the characteristics of depression-related neural variability using RCMPE with resting-state electroencephalogram of patients with major depressive disorder (MDD) and healthy controls. We found that neural variability in the coarse temporal scale of MDD rose in the overall area and that the severity of depressive symptoms was inversely proportional to neural variability in the fine temporal scale of the frontal area. Furthermore, these results differed from characteristics observed with the conventional signal or frequency analysis method. Moreover, features of neural variability could play different roles depending on the temporal scale. Based on the results of our first study, we hypothesized that RCMPE would reveal novel neural correlates related to neural variability and cognitive processing. Accordingly, the aim of the second study was to confirm changes in neuronal variability during the feedback process. We applied a new analysis method called event-related variability, which combines RCMPE and the sliding window method. In addition, through event-related variability, areas in which neural variability differed between positive and negative feedbacks were identified, and whether or not this variability difference was related to learning performance in the probability learning task was determined. We found that neural variability significantly differed between positive and negative feedbacks in several areas, including the frontal and occipital areas on a fine temporal scale and the central area on a coarse temporal scale. In addition, learning performance of the probabilistic learning task was directly proportional to the degree of fine variability in the frontal area between the two feedbacks. Compared to the event-related potential, which is a conventional measure of cognitive processing, in the temporal and spatial dimensions, some characteristics of event-related variability differed. Through both studies, neuronal variability is not only related to the pathophysiology of MDD but also to the feedback process in learning mechanism. In addition, new analysis methods, such as RCMPE and event-induced variability, could help find a new measure of neural variability related to neuropsychiatric disorders and cognitive function.