Indications for network regularization during absence seizures: Weighted and unweighted graph theoretical analyses

https://doi.org/10.1016/j.expneurol.2009.02.001Get rights and content

Abstract

Previous studies with intracranial recordings suggested that a more random spatial structure of functional brain networks could be related to seizure generation. Here, we studied whether similar network changes in weighted and unweighted networks can be found in generalized absence seizures recorded with surface EEG. We retrospectively selected EEG recordings of eleven children with absence seizures. The functional neural networks were characterized by calculating both coherence and synchronization likelihood (SL) between 21 EEG signals that were either broad band filtered (1–48 Hz) or filtered in different frequency bands. From both weighted and unweighted networks the clustering coefficient (C) and path length (L) were computed and compared to 500 random networks. We compared the ictal with the pre-ictal network structure. During absence seizures there was an increase of synchronization in all frequency bands, seen most clearly in the SL-based networks, and the functional network topology changed towards a more ordered pattern, with an increase of C/C-s and L/L-s. This study supports the hypothesis of functional neural network changes during absence seizures. The network became more regularized in weighted and unweighted analyses, when compared to the more randomized pre-ictal network configuration.

Introduction

Epilepsy is a paroxysmal disorder of the brain, characterized by a sudden and unpredictable occurrence of seizures. To increase the knowledge of seizure generation, much research has been done on seizure dynamics and changes in synchronization between cortical areas (Le Van Quyen et al., 1998, Guye et al., 2006, Schindler et al., 2007). In this study we focused on absence seizures, a frequent, well-defined pattern of generalized seizures observed in childhood or juvenile absence epilepsies (Sadleir et al., 2006). Typical absence seizures in this context coincides with EEG features of generalized bilateral synchronous 3-Hz spike wave discharges with a sudden on- and offset (Commission on Classification and Terminology of the International League Against Epilepsy, 1989). The EEG interictal background activity is almost always normal. We used these typical seizures to study changes of the functional neural network during seizure activity.

Synchronization of neural activity in the brain is considered to be essential for information processing, but may also be an important factor in seizure dynamics (Uhlhaas and Singer, 2006). Both in model studies and functional network studies, changes in synchronization during and before seizures have been investigated. Seizures are often characterized as ‘hypersynchronous states’, but several studies showed that this description is an oversimplification of the synchronization process during seizures with various etiology (Wendling et al., 2005, Guye et al., 2006, Schindler et al., 2007). Regarding absence seizures, Aarabi et al. (2008) studied synchronization levels prior to and during absence seizures. An increase of synchronization during the seizures was detected, whereas the pre-ictal state did not show similar changes in all patients.

Our goal in this study was to explore the behavior of the functional neural network, as determined by patterns of pair-wise synchronization between EEG channels, during absence seizures. We used graph theory to provide a model of the neural network. In 1998, Watts and Strogatz (1998) introduced the concept of characterizing networks by their local clustering and overall connectedness. Optimally functioning networks (so called small-world networks) have a high local clustering and a few long-distance connections, resulting in high overall connectedness. Since then, in both neuro-anatomical and functional network studies, network analysis has been applied, and it has shown small-world features in healthy neural networks (Sporns and Zwi, 2004, Bassett and Bullmore, 2006, He et al., 2007). Indications exist that in brain diseases (e.g. epilepsy, M. Alzheimer, brain tumors), the functional network has been damaged and therefore displays a less optimal spatial organization (Bartolomei et al., 2006, Ponten et al., 2007, Stam et al., 2007a, Douw et al., 2008). A few recent hippocampal slice-model and in vitro studies have tried to characterize the neural network during seizing, and we might conclude that changes in network structure play an important role in seizure emergence (Netoff et al., 2004, Percha et al., 2005, Dyhrfjeld-Johnsen et al., 2007, Srinivas et al., 2007, Morgan and Soltesz, 2008).

The first study concerning this topic showed that networks with short path-lengths (small-world or random networks) synchronize more easily, and therefore might be more vulnerable to seizures (Netoff et al., 2004). Seizing started with the drop of recurrent connections and enhancement of the synaptic strength. During seizures, the network corresponded to a small-world regimen, and the clustering coefficient dropped at the beginning of bursting, which signaled the transition towards a more random regimen. This suggested that bursting behavior may represent a dynamical state beyond seizures. The basic principle in another study was that epileptogenesis in temporal lobe epilepsy is characterized by structural network remodeling and axonal sprouting. A model of coupled non-identical neurons was used, to explore network characteristics by increasing the rewiring probability p. The authors found an abrupt transition from a disordered to a globally ordered state when increasing p (Percha et al., 2005). This transition might play an important role in seizure emergence, as one of the underlying mechanisms of temporal epilepsy is sprouting. Another study used a large computational network model, in which they gradually removed the long-distance connections simulating sclerosis (Dyhrfjeld-Johnsen et al., 2007). During this process, small-world characteristics increased together with enlarged hyperexcitability, as long as a few long distant connections were preserved. This might implicate that after brain damage a few lasting neurons and fibers can cause hyperexcitability or seizures. They also examined the role of microcircuits and highly interconnected hubs in hyperexcitability (Morgan and Soltesz, 2008). Well connected hubs in networks may play an important role in seizure genesis. Another in vitro study used injured hippocampal neurons, in which the neural network became hypersynchronous and fired bursts at high frequency, described as ‘induced epileptic activity’ (Srinivas et al., 2007). The neural network became more random after the injury, with a decreased clustering coefficient.

In a previous study we have performed network analysis on intracranial electroencephalogram (EEG) recordings from focal seizures in mesial temporal lobe epilepsy patients. We have found a more random interictal functional network organization compared to the ictal network structure (Ponten et al., 2007). Our finding that during the seizure the functional network changes, is supported by two other clinical studies and a single patient study (Wu et al., 2006, Kramer et al., 2008, Schindler et al., 2008). The single patient study showed an increase of the clustering coefficient during the seizure compared to the pre-ictal state, consistent with a less random-like network (Wu et al., 2006). A recent study, using intracranial EEG recordings, focused on seizure onset in relation to network topology (Kramer et al., 2008). These authors also concluded that the emergent coupling between the electrodes changes at seizure onset, with a decrease of randomness during the seizures. The most recent study also observed a relative shift from a random towards a more regular functional topology during intracranial recorded focal seizures (Schindler et al., 2008). These changes were accompanied by a decreased stability of the globally synchronized state during the seizures, which increased already prior to seizure end.

For further exploration of the application of network analysis in seizures, we performed the present study. We selected standard EEG recordings from eleven patients with absence seizures. The main purpose of this study was to evaluate the hypothesis that changes of the functional network occur during generalized absence seizures, compared to the pre-ictal network properties. Based on intracranial recordings of focal temporal lobe seizures, our hypothesis was for networks to become more regular during seizures. Apart from this main question, we were also interested in some methodological issues. Firstly, we investigated what type of analysis would be best to distinguish ictal and pre-ictal dynamics: linear or nonlinear based synchronization characterization of the network. Secondly, to extend our knowledge of functional neural networks, we wanted to know whether unweighted and weighted network analyses are equally useful for this purpose. As all previously mentioned studies used intracranial recordings, and patients suffering from absence epilepsy never undergo intracranial recording, we also investigated whether surface EEG is suitable to distinguish seizures from non-seizure activity, based on network features.

Section snippets

Patient selection and EEG recording

We retrospectively selected 11 children with absence seizures during EEG recording, using the EEG database of the department of Clinical Neurophysiology at the VU University Medical Center (Amsterdam, the Netherlands). Patient characteristics are shown in Table 1. All these patients fulfilled the criteria of CAE (Commission on Classification and Terminology of the International League Against Epilepsy, 1989). The recordings were performed in the context of routine medical care and were analyzed

Synchronization Likelihood (SL)

The mean SL values in the different epochs, broad band filtered and filtered in the frequency bands (delta 1–4 Hz, theta 4–8 Hz, alpha 8–13 Hz, beta 13–30 Hz, and gamma 30–48 Hz) were calculated. These results were normalized by dividing them by the value of the first epoch, and are shown in Fig. 1a. We found a significant increase (between two and almost nine times higher) of the mean SL during the seizures compared to the pre-ictal state, both in the broad frequency band and in the filtered

Discussion

In this study, we were the first to perform network analysis with surface EEG recordings during generalized absence seizures. As expected, we found an increase of synchronization (linear and nonlinear) during the seizure, compared to the pre-ictal state. Of more interest are the network changes we found, namely an increase of the clustering coefficient (C/C-s) and path-length (L/L-s; weighted and unweighted), indicating a more regular organization during absences. Where previous studies used

Disclosure

The authors report no conflicts of interest.

Acknowledgments

Els van Deventer is thanked for the thorough literature search.

Funding: This work is financially supported by the Dutch National Epilepsy Fund (NEF) (05-12 to SCP and 08-08 to LD).

References (36)

  • ReijneveldJ.C. et al.

    The application of graph theoretical analysis to complex networks in the brain

    Clin. Neurophysiol.

    (2007)
  • StamC.J. et al.

    Synchronization likelihood: an unbiased measure of generalized synchronization in mulitvariate data sets

    Physica, D

    (2002)
  • UhlhaasP.J. et al.

    Neural synchrony in brain disorders: relevance for cognitive dysfunctions and pathophysiology

    Neuron

    (2006)
  • WinterhalderM. et al.

    Sensitivity and specificity of coherence and phase synchronization analysis

    Phys. Lett., A

    (2006)
  • BassettD.S. et al.

    Small-world brain networks

    Neuroscientist

    (2006)
  • ChavezM. et al.

    Synchronizing weighted complex networks

    Chaos: An Interdisciplinary J. Nonlinear Sci.

    (2006)
  • Proposal for revised classification of epilepsies and epileptic syndromes. Commission on Classification and Terminology of the International League Against Epilepsy

    Epilepsia

    (1989)
  • Dyhrfjeld-JohnsenJ. et al.

    Topological determinants of epileptogenesis in large-scale structural and functional models of the dentate gyrus derived from experimental data

    J. Neurophysiol.

    (2007)
  • Cited by (129)

    View all citing articles on Scopus
    View full text