In this contribution a new algorithm based on the spatio-temporal dynamics of Reaction-Diffusion Cellular Nonlinear Networks (RD-CNN) for analyzing brain electrical activity in epilepsy is proposed. RD-CNN are determined in an identification process and then analyzed by means of Chuas Local Activity theory [1]. Clinical manifestations of epileptic seizures are phenomena of abnormal, excessive, or synchronous neuronal activity in the brain. The epileptic disorder is the most common chronic disorder of the central nervous system. Generally seizures appear without sign or warning and the problem of detecting a possible pre-seizure state in epilepsy from electroencephalogram (EEG) time series analysis has already been addressed in many publications. Up to now the identification of an impending epileptic seizure with sufficient specificity and reliability cannot be performed in an automated procedure. An algorithm for a reliable prediction of epileptic seizures would enable the realization of implantable seizure warning devices, which could provide valuable information to the patient and time/event specific drug delivery or direct electric nerve stimulation. CNN are characterized by local couplings of comparatively simple dynamical systems. With this property these networks are well suited to be realized as parallel, analog computing architectures. Recently developed CNN hardware realizations exhibit high parallel computing capacity together with low power consumption. Thus CNN platforms could provide a basis for future implantable seizure warning and preventing devices. In the proposed procedure, RD-CNN are considered for an identification of consecutive segments of intracranial EEG using supervised parameter optimization procedure. These networks are deduced from Reaction-Diffusion Systems, which are applied in different fields of science to describe various complex phenomena like nonlinear wave propagation or pattern formation. Especially, Chua [1] derived the Local Activity Theory, which provides a quantitative, necessary condition for emergent behavior in RD-CNN. The aim of these first investigations is to apply the Local Activity Theory to the networks obtained in the proposed identification procedure. Results for long time recordings gained during presurgical diagnostics in temporal lobe epilepsy are presented.