Wavelet transforms for electroencephalographic spike and seizure detection

The application of wavelet transforms (WT) to experimental data from the nervous system has been hindered by the lack of a straightforward method to handle noise. A noise reduction technique, developed recently for use in wavelet cluster analysis in cosmology and astronomy, is here adapted for electroencephalographic (EEG) time-series data. Noise is filtered using control surrogate data sets generated from randomized aspects of the original time-series. In this study, WT were applied to EEG data from human patients undergoing brain mapping with implanted subdural electrodes for the localization of epileptic seizure foci. EEG data in 1D were analyzed from individual electrodes, and 2D data from electrode grids. These techniques are a powerful means to identify epileptic spikes in such data, and offer a method to identity the onset and spatial extent of epileptic seizure foci. The method is readily applied to the detection of structure in stationary and non-stationary time-series from a variety of physical systems.