Epilepsy is the most common chronic disorder of the nervous system. Generally, epileptic seizures appear without foregoing sign or warning. The problem of detecting a possible pre-seizure state in epilepsy from EEG signals has been addressed by many authors over the past decades. Different approaches of time series analysis of brain electrical activity already are providing valuable insights into the underlying complex dynamics. But the main goal the identification of an impending epileptic seizure with a sufficient specificity and reliability, has not been achieved up to now.
An algorithm for a reliable, automated 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 possibly a direct electrical nerve stimulation. Cellular Nonlinear Networks (CNN) are promising candidates for future seizure warning devices. CNN are characterized by local couplings of comparatively simple dynamical systems. With this property these networks are well suited to be realized as highly parallel, analog computer chips. Today available CNN hardware realizations exhibit a processing speed in the range of TeraOps combined with low power consumption.
In this contribution new algorithms based on the spatio-temporal dynamics of CNN are considered in order to analyze intracranial EEG signals and thus taking into account mutual dependencies between neighboring regions of the brain. In an identification procedure Reaction-Diffusion CNN (RD-CNN) are determined for short segments of brain electrical activity, by means of a supervised parameter optimization. RD-CNN are deduced from Reaction-Diffusion Systems, which usually are applied to investigate complex phenomena like nonlinear wave propagation or pattern formation. The Local Activity Theory provides a necessary condition for emergent behavior in RD-CNN. In comparison linear spatio-temporal autoregressive filter models are considered, for a prediction of EEG signal values. Thus Signal features values for successive, short, quasi stationary segments of brain electrical activity can be obtained, with the objective of detecting distinct changes prior to impending epileptic seizures.
Furthermore long term recordings gained during presurgical diagnostics in temporal lobe epilepsy are analyzed and the predictive performance of the extracted features is evaluated statistically. Therefore a Receiver Operating Characteristic analysis is considered, assessing the distinguishability between distributions of supposed preictal and interictal periods.
In previous publications,1-6 several approaches targeting the problem of seizure prediction7 in epilepsy8 have been
proposed. In this contribution recent results based on an EEG-signal prediction algorithm will be presented and
discussed in detail. Therefore segmented data aquired by multi-electrode Stereoelectroencephalography (SEEG)
and Electrocorticography (ECoG) are presented to a delay-type DTCNN with linear weight functions and a 3×1
network topology. This leads to series of signal predictors and according to that to series of prediction errors.
These prediction error series are arranged in a 2 dimensional representation called error profile.9 This profile
enables the choice of optimal positions for implanting long time electrodes, by means of which perhaps a mostly
effective seizure prediction may become possible. So far data of different patients have been studied in detail
and some distinct electrode points were found showing distinct changes before a seizure onset.
Reaction-Diffusion systems can be applied to describe a broad class of nonlinear phenomena, in particular in biological systems and in the propagation of nonlinear waves in excitable media. Especially, pattern formation and chaotic behavior are observed in Reaction-Diffusion systems and can be analyzed. Due to their structure multi-layer Cellular Neural Networks (CNN) are capable of representing Reaction-Diffusion systems effectively. In this contribution Reaction-Diffusion CNN are considered for modeling dynamics of brain activity in epilepsy. Thereby the parameters of Reaction-Diffusion systems are determined in a supervised optimization process, and brain electrical activity using invasive multi-electrode EEG recordings is analyzed with the aim to detect of precursors of impending epileptic seizures. A detailed discussion of first results and potentiality of the proposed approach will be given.