Networked information interactions of epileptic EEG based on symbolic transfer entropy

Abstract Identifying networked information exchanges among brain regions is important for understanding the brain structure. We employ symbolic transfer entropy to facilitate the construction of networked information interactions for EEGs of 22 epileptics and 22 healthy subjects. The epileptic patients during seizure-free interval have lower information transfer in each individual and whole brain regions than the healthy subjects. Among all of the brain regions, the information flows out of and into the brain area of O1 of the epileptic EEGs are significantly lower than those of the healthy (p<0.0005), and the information flow from F7 to F8 (p<0.00001) is particularly promising to discriminate the two groups of EEGs. Moreover, Shannon entropy of probability distributions of information exchanges suggests that the healthy EEGs have higher complexity and irregularity than the epileptic brain electrical activities. By characterizing the brain networked information interactions, our findings highlight the long-term reduced information exchanges, degree of brain interactivities and informational complexity of the epileptic EEG.