Purpose: An understanding of the principles governing the behavior of
complex neuronal networks, in particular their capability of generatin
g epileptic seizures implies the characterization of the conditions un
der which a transition from the interictal to the ictal state takes pl
ace. Signal analysis methods derived from the theory of nonlinear dyna
mics provide new tools to characterize the behavior of such networks,
and are particularly relevant for the analysis of epileptiform activit
y. Methods: We calculated the correlation dimension, tested for irreve
rsibility, and made recurrence plots of EEG signals recorded intracran
ially both during interictal and ictal states in temporal lobe epileps
y patients who were surgical candidates. Results: Epileptic seizure ac
tivity often, but not always, emerges as a low-dimensional oscillation
. In general, the seizure behaves as a nonstationary phenomenon during
which both phases of low and high complexity may occur. Nevertheless
a low dimension may be found mainly in the zone of ictal onset and nea
rby structures. Both the zone of ictal onset and the pattern of propag
ation of seizure activity in the brain could be identified using this
type of analysis. Furthermore, the results obtained were in close agre
ement with visual inspection of the EEG records. Conclusions: Applicat
ion of these mathematical tools provides novel insights into the spati
o-temporal dynamics of ''epileptic brain states''. In this way it may
be of practical use in the localization of an epileptogenic region in
the brain, and thus be of assistance in the presurgical evaluation of
patients with localization-related epilepsy.