Rationale:
When perturbation in a dynamical system is not efficiently dissipated, it can enable the system to explore its state space (Deco et al., 2011) and in the case of an epileptic system, it can initiate and spread the epileptiform activity (Suffczynski et al., 2006; Stacey et al., 2011). In many computational models of epilepsy, the sudden and unexpected seizure initiation originates from the interplay of perturbations with the underlying dynamical system (Jirsa et al. 2014). However, these notions have not been given a direct interpretation in the analysis of the epileptic functional connectivity networks. Here, we study the stability of these networks by exploring their response to perturbing events.
Method:
We simulated the epileptic networks based on a simple concept from fluid mechanics and computed the speed of perturbation dissipation during the ictal, pre- and post-ictal periods. Furthermore, by recruiting two graph theory measures (Betweennes Centrality and Small-Worldness), we explored two relevant topological features of these networks.
Results:
Our results indicate slower perturbation dissipation speed prior to seizure initiation compared to during and after the seizure (p=9.56e-10 and p=6.36e-4 respectively). The lower dissipation speed and consequently longer dissipation time can lead to noise accumulation in the network, potentially causing a sudden relaxation through cascading failure in the form of seizures. Furthermore, we show that, compared to other periods, the periods with lower perturbation dissipation speed were accompanied by significantly higher node loads (e.g. during the pre-ictal period with p< 0.001). And the periods with higher perturbation dissipation speed co-occurred with higher Small-Worldness (e.g. during the ictal period with p< 0.001).
Conclusion:
Cumulatively, our results suggest disturbed relaxation, higher noise accumulation, and higher node loads before seizure onset and increased capability for dissipating perturbation during seizures through network topological reconfiguration.
These findings advance our understanding of the underlying mechanism of ictal activity. Yet, this framework can be further extended to rank different brain regions in terms of their impact on the network stability with a potential implication in identifying the critical regions initiating and spreading seizures.
Funding:
:This research was supported by CIHR project grant (390044); Canada Research Chair Program (DKN); NSERC Discovery Grant RGPIN-2014-06089 (PP).
FIGURES
Figure 1