Seizure prediction by non-linear time series analysis of brain electrical activity.

Brain electrical activity of 16 patients with temporal lobe epilepsy, recorded intracranially during seizure-free intervals as well as during transitions to the seizure state, was analysed using methods derived from the theory of non-linear dynamics. Long-lasting and marked changes towards low-dimensional system states were found to occur specifically up to 25 min prior to epileptic seizures and allow to predict the occurrence of individual seizures in time. These findings reflect a continuous increase in the degree of synchronicity, and thus open a window for the study of mechanisms generating seizures in humans. This offers new possibilities for therapeutic interventions.