Abstracts

Rationale: For over a century, extensive documentation has shown that seizures tend to cluster in time. However, the physiological mechanisms underlying seizure clusters remain poorly understood. Models that quantify long-term temporal patterns of epileptic activity will improve clinical management and guide predictive strategies. Methods: We investigated clustering of interictal spikes, subclinical seizures, and seizures in 15 subjects with drug-resistant, focal epilepsy. Data was obtained during a clinical trial (between 6 months to 3 years duration) for an implantable seizure prediction device [1]. Hundreds of thousands of interictal spikes and up to thousands of seizures were recorded [1,2]. We postulated that clustered seizures arise naturally from a long-range dependent process, where correlations between consecutive inter-seizure intervals result in periods of high activity as well as long periods of quiescence. To investigate this hypothesis, we measured the Fano factor (ratio of variance to mean) of the event rate over multiple temporal windows (minutes to days for spikes, and days to months for seizures). We also measured the Hurst exponent for the inter-event intervals. We have previously used the Hurst exponent to establish long-range dependence in seizure timing [3], and here we applied this measure to interictal spike times. Results: Fig. 1 shows that Fano factors were much greater than 1 for interictal spike rate, and also for a majority of subjects' seizure rate. This result indicates extreme overdispersion, with large deviations from the mean event frequency, as opposed to a Poisson process, where the Fano factor is equal to 1. Growth of the Fano factor with increasing window size, as observed in Fig. 1, is consistent with long-range dependence in a time series [4]. Indeed, the interictal spike intervals for all 15 subjects showed a Hurst exponent greater than 0.5 (ranging from 0.71 to 0.83). An exponent of 0.5 indicates a random, or memoryless process, whereas values between 0.5 and 1 reflect a correlation structure where short intervals follow short intervals (as in clustering) and vice versa. Conclusions: Multiple scales of epileptic activity, from spike-waves to clinical seizures, demonstrate hallmarks of long-range dependence in their timing. The ubiquity of these phenomena across different event types suggests that system memory is a fundamental property of epileptic networks. In a clinical context, analysis of clustered seizures should not be divorced from isolated events, as both clusters and unusually long seizure-free periods can arise from a single long-memory process. Modelling seizure variability on an individual level is vital to correctly evaluate a response to treatment. References 1. Cook, M., et al. "Prediction of seizure likelihood with a long-term, implanted seizure advisory system in patients with drug-resistant epilepsy: a first-in-man study" Lancet Neurol. 12.6 (2013): 563-571. 2. Karoly, P., et al. "Interictal spikes and epileptic seizures: their relationship and underlying rhythmicity" Brain (2016) 3. Cook, M., et al. "The Dynamics of the Epileptic Brain Reveal Long-Memory Processes" Front. Neurol. 5.217 (2014) 4. Abry, P., et al. "Wavelets for the analysis, estimation and synthesis of scaling data" Self-similar network traffic and performance evaluation, pp. 39-88 Park and Willinger, eds. New York: Wiley, 2000. Funding: This project was funded by an Australian National Health and Medical Research Council Project Grant (APP1065638)