ABSTRACT
Analysis techniques designed to reliably predict epileptic seizures from recordings of brain electrical activity could make a considerable contribution to improving the quality of life of our patients by providing newer and more effective treatments and expanding our knowledge of how seizures are generated in the brain. While research in the 1990s mainly focused on a characterization of seizure precursors using univariate time series analysis techniques, recent years have seen a shift toward a characterization of interdependencies between different brain regions during the preictal period. For an overview, see the works by Mormann et al. (2007) and Schelter et al. (2008). In addition to bivariate time series analysis techniques that aim at estimating the strength and direction of an interaction (Lehnertz et al. 2009), the success of network theory in physics, biology, and other scienti–c –elds (Boccaletti et al. 2006; Arenas et al. 2008) has inspired the development of multivariate analysis techniques. These methods aim at characterizing the interdependence structure between arbitrary numbers of interacting dynamical systems.
