Polytopic linear parameter-varying model of epileptiform activity

A novel abstract modeling approach for brain dynamics exhibiting epileptiform activity is proposed, and a seizure prediction algorithm based on this approach is presented. The model consists of several polytopes of parameters, each of which corresponds to a particular brain dynamics, and a seizure onset is predicted when real time-identified model parameters make a transition from an interictal polytope to an ictal polytope in a probabilistic sense. The seizure-prediction capability of the developed algorithm is assessed using simulated EEG signals.