Abstracts

Rationale: Previous studies provided first evidence that an improved detection of pre-ictal states can be achieved through a quantification of mutual relationships between different functional subsystems that compose an epileptic brain network. We here investigate how weighted and directed short- and long-ranged interactions influence the ability to differentiate between inter-ictal and pre-ictal states. Methods: We analyzed intracranial multi-channel (mean: 54 sites), multi-day (on average 110 hours per patient) EEG recordings from up to now 15 epilepsy patients (8 with unifocal and 7 with multifocal epilepsies; all patients are seizure-free post-operatively). A total of 68 seizures occurred. Between the clinically defined epileptic focus and other sampled brain regions the strength and the direction of interactions were estimated in a time-resolved manner each via a phase-based (strength: R, direction: D) and an information theoretical approach (strength: G, direction: T). Of these interactions those between the focus and its vicinity were defined as short-ranged. Results: We observed the distributions of measures for the pre-ictal (assumed duration 4 hours) and the inter-ictal state (remaining data, excluding seizures and a post-ictal period of 30 minutes) to differ significantly, at least in a number of patients. However, differentiation depended on daily rhythms and on the approach used to characterize interaction properties (day: R 67%, D 33%, G 42%, T 67%; night: R 58%, D 58%, G 33%, T 75% of cases). Moreover, separability of states appeared to be influenced by contributions of short- and long-ranged interactions, which differently affected strength and direction of interactions. Conclusions: Our findings indicate that considering differentially the contributions of weighted and directed short- and long-ranged interactions can improve the ability to distinguish between pre-ictal and inter-ictal states. This approach could improve our understanding of ictogenesis in epileptic brain networks. This work was supported by the Deutsche Forschungsgemeinschaft (Grant No: LE 660/5-2)