PREDICTABILITY OF SEIZURE-LIKE EVENTS IN A COMPLEX NETWORK MODEL OF INTEGRATE-AND-FIRE NEURONS

Prediction of epileptic seizures relies on a consistent relationship between indices of brain activity on the one hand and seizure probabilities on the other hand. We study a qualitative complex network model of integrate-and-fire neurons which generates both asynchronous behavior and short seizure-like events of collective firing. The network model is interpreted as a small seizure-generating zone, which is influenced by various endogenous and exogenous factors, which possibly modulate seizure probability and which are represented by model parameters. For constant parameters the events seem unpredictable, and we find that an increase in seizure probability is accompanied by an increase of fluctuations of the asynchronous behavior, for various considered influencing factors. Using this observation, we find some predictability if the model is endowed with randomly varying model parameters, which could reflect changes in endogenous and exogenous factors that might influence ictogenesis.