EEG analysis by multi layer Cellular Nonlinear Networks (CNN)

The analyses of EEG-signals of patients suffering from epilepsy have been performed by many authors during the last years. The main goal of these analyses is to enable a detection of seizure precursors. Several methods based on CNN - e.g. the approximation of the correlation dimension, the prediction of EEG-signals, the pattern detection algorithm - have been proposed and studied in detail. Yielding interesting results, the signal prediction algorithm has been analyzed in more detail in order to optimize the obtained results of the predictor system, both for quality and computational complexity. Applying a CNN predictor to recordings of multi EEG electrodes results in a so called prediction error profile. Electrodes which show the most significant changes before epileptic seizures could be identified by using these profiles.