Abstracts

We have previously showed that an increase of the relative phase clustering index (rPCI) in the gamma frequency band (30-120Hz) of EEG and MEG recordings is associated with a transition between interictal and ictal neuronal behaviour in patients with photosensitive epilepsy (Kalitzin et al. 2002, Parra et al 2003). Peri-ictal SEEG data in temporal lobe epilepsy show that high rPCI values are indicative of the seizure onset site. In addition a further increase of rPCI extensive to neighbouring recording sites suggests increased probability for seizure occurrence (Velis et al 2002). In this contribution we provide further insight into the mechanisms of rPCI from the perspective of artificial neuronal modelling.
We use a cortico-thalamic model of bi-stable neuronal dynamics developed previously by our group. The model is built within Simulink/Matlab environment. In addition, we have applied analytical method of complex transfer functions to explore linearised properties of the model and to compute rPCI values for different model parameters.
Using simulations and analytical tools the following properties of our model were elicited:
Our cortico-thalamic model has a pair of dynamic phase space attractors for a wide range of choice of the model parameters. One of the attractors corresponds to normal state and the second represents a limit cycle that can model ictal activity.
The probability of the transition from normal to ictal activity can be controlled by the model parameters:it increases with the amount of noise in the cortical circuit.
The introduction of additional fast inhibitory synapses in the population of cortical interneurons leads to gamma-frequency resonance properties that may account for the appearance of rPCI at those frequencies. Analytic conditions for such a scenario have been obtained and validated through simulations.
rPCI values can increase for higher values of the noise in the cortical network and, therefore, this quantity can indirectly represent a higher probability for ictal transition.
rPCI can be also enhanced by [quot]sharper[quot] resonance properties of the network, which in turn may also facilitate ictal onsets.
General analytic techniques and model simulations show that stimulation paradigms consisting of short pulses with low duty cycle can generate higher rPCI in a system with resonant properties.
Seizures in bi-stable neural networks cannot be predicted because they are triggered by random fluctuations. Nevertheless, the instantaneous probability of ictal transition can be correlated to the rPCI generated by a cortico-thalamic model neural network. This finding provides a plausible explanation of our empirical observations in EEG signals of epileptic patients. On the basis of our model analysis we can derive an optimal stimulation paradigm for calculating the ictal onset probability and we have applied this during video/SEEG seizure monitoring in temporal lobe epilepsy