Authors :
Elaheh Zarafshan, Ms. – Florida International University; Parisa Forouzannezhad, PhD – University of Washington; Robin Perry Mayrand, Ms. – Florida International University; Mercedes Cabrerizo, PhD – Florida International University; Alberto Pinzon, MD – Epilepsy Center, Baptist Hospital of Miam; Malek Adjouadi, PhD – Florida International University
Rationale: EEG-signal generated by the brain is highly nonlinear and non-stationary making non-linear-based methods more suitable and more fitting for functional connectivity networks (FCNs)-based calculations and rendering [1]. An information theoretic and non-parametric approach, such as mutual information (MI), could be used to detect and quantify the non-directional dependency between different time series [2]. We applied this technique to data obtained from the scalp EEG of four adults recorded at Baptist Hospital in Miami, Florida.
Methods: We examined the FCNs of different interictal epileptiform discharges (ED: single, complex, and repetitive spikes) in 4 different regions of the brain between electrode pairs in one-second segments. As shown in Figure 1 (panel A), forty-five segments were chosen randomly for the ED types from the 19 electrodes using the standard 10-20 EEG system. Using the MI approach with bin=10, we calculated connectivity matrices that help identify network synchronizations based on the number of connections and strength of connections for all brain regions. Next, to calculate neuronal networks with a so-called "small-world" topography [3], the connectivity matrices were assumed as a weighted undirected adjacency matrix of a graph in which nodes are electrodes and edges are connectivity strength, representing links between two associated electrodes.
Results: The neuronal
networks topography of three ED types resulted in nearly a random network, meaning that the small-world characteristic is lost due to interictal abnormal activity (Figure 1, panel B). The activities in four regions of the brain resulted in different connectivity patterns in the three ED types, including within and between brain regions. As shown in Figure 2, panel D, compared to the complex and repetitive spikes, a lower number of connections with weaker connectivity strength were observed for the single spike in all regions. These findings underscore the importance of exploring different ED types in EEG data. For instance, based on our findings, while single spike cases are more likely indicative of focal epilepsy, the complex and repetitive spikes are indicative of generalized epilepsy, with stronger and more prevalent connections in all regions of the brain. The one-way ANOVA test showed statistically significant different between the average connectivity of these three ED types (p-values < .05).