State transfer network of time series based on visibility graph analysis for classifying and prediction of epilepsy seizures

Visibility graph analysis of time series became widely used as a time series analysis in the recent years. State transfer network is a network of mapping mono/multivariate time series into a network of local states based on visibility graph, it was used to study the evolutionary behavior of time series and in this study, we applied this principle to the detection of epileptic seizures. Two sets of EEG data were used; first set was obtained from subjects with the healthy brain and the second set obtained from an unhealthy part of the brain during existence of epileptic seizures. Results show a clear discrepancy between the two groups of data with a dominantly appearance of particular nodes in the networks of an epileptic group called hubs or motif, accordingly, the visibility graph network analysis based analysis can be considered as a prediction way of epilepsy seizures.