Synchronization phenomena in human epileptic brain networks

Abstract

Epilepsy is a malfunction of the brain that affects over 50 million people worldwide. Epileptic seizures are usually characterized by an abnormal synchronized firing of neurons involved in the epileptic process. In human epilepsy the exact mechanisms underlying seizure generation are still uncertain as are mechanisms underlying seizure spreading and termination. There is now growing evidence that an improved understanding of the epileptic process can be achieved through the analysis of properties of epileptic brain networks and through the analysis of interactions in such networks. In this overview, we summarize recent methodological developments to assess synchronization phenomena in human epileptic brain networks and present findings obtained from analyses of brain electromagnetic signals recorded in epilepsy patients.

Introduction

The term synchronization is derived from the Greek words $\sigma \upsilon \nu$ ($\text{syn}=\text{common}$, together) and $\chi \rho o\nu o\zeta$ ($\text{chronos}=\text{time}$), meaning shared or common time. The notion of two interacting systems oscillating at the same time was introduced to classical physics by the Dutch mathematician, astronomer, and physicist Christiaan Huygens (Huygens, 1673). Today, the study of synchronization of coupled dynamical systems is an active field of research in many scientific and technical disciplines Pikovsky et al., 2001, Boccaletti et al., 2002. Synchronization phenomena can manifest themselves in many different ways, and in many instances they may appear quite counter-intuitive. Thus, over the past two decades different concepts of synchronization have been proposed. Apart from the simple case of complete synchronization, where the state variables become identical for large times, the term lag synchronization can be used to describe a state where the state variables of two interacting systems coincide if one is shifted in time. Furthermore, the classical concept of phase synchronization, as proposed by Huygens, was extended from linear to nonlinear or even chaotic systems for cases where the definition of a phase variable is possible for the systems under investigation. The concept of generalized synchronization was introduced for driver–response systems to describe a state in which the state variables of the systems are connected by a functional relationship with certain mathematical properties. With the notion of interdependence one does not even assume that a functional relationship exists, and similarities of local neighborhoods in the state spaces of the systems are exploited to derive quantifying criteria. Due to the diversity of observable synchronization phenomena attempts have been made to find a unifying framework for synchronization of coupled dynamical systems Brown and Kocarev, 2000, Boccaletti et al., 2001, Hramov and Koronovskii, 2005.

The human brain is a complex network of interacting subsystems, and it is now commonly accepted that synchronization plays an important role in brain functioning and dysfunctioning Varela et al., 2001, Glass, 2001, Schnitzler and Gross, 2005, Uhlhaas and Singer, 2006, Buzsáki, 2006. A prominent example for pathophysiologic neuronal synchronization is epilepsy along with its cardinal symptom, the epileptic seizure. Epilepsy affects approximately 1% of the world’s population (Duncan et al., 2006). Two-thirds of affected individuals have seizures that can be sufficiently controlled by antiepileptic drugs (Kwan and Brodie, 2006), and another 7–8% may profit from epilepsy surgery (Duncan, 2007). In about 25% of individuals with epilepsy, however, seizures cannot be controlled by any available therapy.

Despite the many improvements in understanding the complicated disease epilepsy (Engel and Pedley, 2007) there exist a number of problems for which there are currently no satisfactory solutions. In clinical epileptology, the unequivocal identification of the seizure-generating brain structure (usually referred to as epileptic focus or epileptogenic zone; cf. Rosenow and Lüders, 2001, Kahane et al., 2006) is regarded as a prerequisite for successful surgical treatment. Clinical and anatomic observations together with invasive electroencephalography and functional neuroimaging now provide increasing evidence for the existence of specific cortical and subcortical epileptic networks in the genesis and expression of not only primary generalized but also focal onset seizures Bertram et al., 1998, Bragin et al., 2000, Avoli et al., 2002, Bartolomei et al., 2001, Spencer, 2002, Guye et al., 2006, Bartolomei et al., 2008, Gotman, 2008, Luat and Chugani, 2008. Although it might be intuitively clear to consider seizures as network phenomena, in epileptology, it is still matter of debate, whether the concept of a localized and well-defined epileptic focus should be replaced by that of an epileptic network, whose interactions extend over large regions of the brain. Among others, this controversy can be related to the fact that about 66% of surgically treated patients remain seizure-free (Téllez-Zenteno et al., 2005). Nevertheless, understanding the complex interplay between structure and dynamics of epileptic networks underlying seizure generation could help to improve diagnosis and, more importantly, could advise new treatment strategies, particularly for those patients whose seizures cannot be controlled by any available therapy. The quest for alternative treatment options is also driven by the sudden, unforeseen occurrence of seizures, which represents one of the most disabling aspects of the disease. If it were possible to reliably identify a pre-seizure state, preventive treatment strategies could be replaced by an on-demand therapy in an attempt to reset brain dynamics to a state that will no longer develop into a seizure Lehnertz et al., 2007a, Lehnertz et al., 2007b, Stacey and Litt, 2008. In the field of seizure prediction (see Schelter et al., 2008 for a recent overview), there is now increasing evidence that bivariate analysis techniques – characterizing interactions between different brain regions – appear capable of identifying long-lasting (up to hours) spatial-temporal changes on the ongoing EEG that can be regarded as precursors of an impending focal seizure Mormann et al., 2005, Le Van Quyen et al., 2005, Iasemidis et al., 2005. Some bivariate techniques showed a promising performance that exceeded the chance level as evidenced by statistical validation. The next milestone in the field of seizure prediction is to prove that prediction algorithms can be designed to run prospectively on unselected, out-of-sample data with a performance that is better than that of a random prediction process. In order to achieve this ambitious goal guidelines for designing seizure prediction studies have been proposed (Mormann et al., 2007).

This overview summarizes recent conceptual and methodological developments that aim at improved characterization of interactions in complex networks. It highlights areas that are under active investigation and that promise to provide new insights into the complex spatial-temporal dynamics of human epileptic brain networks.