Abstracts

Rationale: Graph-theoretical approaches to a characterization of anatomical and functional brain networks has been a rapidly evolving field recently. Previous studies on epileptic brain networks reported on an increased regularization of network topology during seizures as compared to the pre- and postictal intervals, and an altered functional brain topology in epilepsy patients can even be observed during the seizure-free interval. However, these studies were based on recordings that lasted from a few seconds to several minutes only. We here investigated the time-course of graph theoretical approaches on the time scales of days particularly with respect to an identification of a pre-ictal state. Methods: We analyzed invasive multi-day, multi-channel EEG recordings from 13 patients with focal epilepsies undergoing pre-surgical evaluation. Using a moving-window approach (duration of each window: 20.48 s corresponding to 4096 data points; no overlap) we estimated the strength of interactions (via mean phase coherence) between all pairs of sampled brain regions. We defined functional network links by thresholding the interaction matrix and estimated the global network characteristics average shortest path length L and clustering coefficient C. Results: Both network characteristics exhibited large fluctuations over time, however with some periodic temporal structure. These fluctuations could -- to a large extent -- be attributed to daily rhythms while relevant aspects of the epileptic process contributed only marginally. Particularly, we could not observe clear cut changes in network states that can be regarded as predictive of an impending seizure. Conclusions: Global statistical properties of epileptic brain networks strongly reflect daily rhythms and possibly alterations of the anticonvulsant medication. Identification of a possible pre-ictal state with graph-theoretical approaches requires a better understanding of these daily rhythms as well as further methodological developments. This work was supported by the Deutsche Forschungsgemeinschaft (Grant No. LE660/4-1)