Abstracts

Rationale: A recent study investigated the statistics of seizure durations in long-term epilepsy data (Cook et al., 2015, Epilepsia). An unusual result from the aforementioned study was that patients with a unimodal distribution of seizure duration performed better than patients with a bimodal distribution in terms of seizure forecast sensitivity in the NeuroVista study (Cook et al., 2013, Lancet Neurology). The link between duration and forecasting accuracy led to the hypothesis that short seizures had different preictal dynamics to long seizures. Here we report on an investigation addressing this hypothesis, providing new fundamental insights into seizure mechanisms. The findings have the potential to change the way seizures are treated, improving patient management. Methods: Continuous ambulatory intracranial EEG data was collected from 15 patients for up to three years, as part of the NeuroVista study (Cook et al., 2013, Lancet Neurology). Seizures were expertly annotated and classified by duration (Cook et al., 2015, Epilepsia). Seizure?'specific neural mass models were created from the data, using machine-learning tools (Freestone and Karoly et al., 2014, Frontiers in Neuroscience). These models provide a time-varying estimate of the effective connectivity of intracortical circuits, which (helps/may help) elucidate mechanisms of seizures. Here we compare the dynamics of effective connectivity over the time course of short and long seizures. Results: We present evidence that seizure durations are governed by the intracortical effective connectivity at seizure onset. Patients who had a unimodal distribution of seizure duration had a single stereotypical connectivity pattern at seizure onset. Patients with a bimodal duration distribution had two distinct, stereotypical connectivity patterns. The differing effective connectivity reflects distinct mechanisms underlying short and long seizures. Conclusions: Short and long seizures are fundamentally different. Patients with multiple seizure types, as classified by duration, may require different therapies for distinct seizure populations. Classification of seizures by duration is relatively simple to implement, yet yields powerful insights that have the potential to revolutionize the treatment of medically refractory epilepsy. Funding: Funded by the National Health and Medical Research Council (NHMRC) of Australia, Project Grant APP1065638.