Application of Quantile Graphs to the Automated Analysis of EEG Signals

Abstract

Epilepsy is classified as a chronic neurological disorder of the brain and affects approximately 2% of the world population. This disorder leads to a reduction in people's productivity and imposes restrictions on their daily lives. Studies of epilepsy often rely on electroencephalogram (EEG) signals to provide information on the behavior of the brain during seizures. Recently, a map from a time series to a network has been proposed and that is based on the concept of transition probabilities; the series results in a so-called quantile graph (QG). Here, this map, which is also called the QG method, is applied for the automatic detection of normal, pre-ictal (preceding a seizure), and ictal (occurring during a seizure) conditions from recorded EEG signals. Our main goal is to illustrate how the differences in dynamics in the EEG signals are reflected in the topology of the corresponding QGs. Based on various network metrics, namely, the clustering coefficient, the shortest path length, the mean jump length, the modularity and the betweenness centrality, our results show that the QG method is able to detect differences in dynamical properties of brain electrical activity from different extracranial and intracranial recording regions and from different physiological and pathological brain states.