Characterization of spatio-temporal changes prior to epileptic seizure by a measure of dynamical similarity

The Anticipation of epileptic seizures has been shown to be feasible with nonlinear mathematical techniques. Although these results are-very promising, further studies are necessary to deeper characterize the dynamics of seizure; transitions. Here we describe a new method to analyze this long-term nonstationarity in the EEG by a measure of dynamical similarity between different parts of the rime series. We apply this method to the study of patients with temporal lobe epilepsy recorded intracranially during transitions to seizure. We show that the method allows to extract spatiotemporal changes in brain dynamics well in advance prior to seizure, offering a new possibility to better understand the pathophysiological mechanism underlying these transitions as well an improved computational strategy.