Abstracts

Rationale: Recent studies demonstrate that epileptic seizure forecasting using intracranial electroencephalographic (iEEG) is possible. Forecasting has the potential to improve the lives of people with epilepsy and may prove useful for adaptive closed-loop therapies.   Methods: Forecasting depends on the hypothesis that there are pre-seizure brain states with increased likelihood of seizure occurrence. Here we assume the pre-seizure state electrical activity is not continuous, but rather intermittently switches in and out of the pre-seizure state.  The iEEG data for this research is wirelessly recorded from intracranial electrodes in dogs and humans with epilepsy (NeuroVista Seizure Advisory System (SAS) or the Medtronic Summit RC+S device). The pre-ictal data was taken from 16-hour periods preceding seizures and the dynamics of pre-ictal events are examined within 16-hour and 4 days prior to seizures in training and test data sets respectively.   Results: Hierarchical clustering was used to separate pre-seizure and inter-ictal states within selected training data segments and logistic regression used to classify test data segments as pre-seizure or interictal. We analyzed the characteristics of the pre-seizure states using 56 training seizures and 41 testing seizures from 4 humans and 4 dogs with epilepsy. Fig. 1 shows pre-ictal and interictal fluctuations followed by a seizure and the distribution function characterizing the pre-seizure states segments distributions within 4 days before seizures across all dogs and humans test data sets which is exponential with R-Squared of 0.48.  Conclusions: The results support a dynamic model of the pre-seizure state characterized by frequent transitions between interictal and pre-seizure that becomes more prevalent within 2 hours before seizure onset. Although the pre-seizure state occurs more frequently as seizures approach onset, the brain switches between normal and pre-seizure states even in the days before seizure.  Funding: This research was supported by Mayo Clinic Discovery Translation Grant, National Institutes of Health (R01-NS063039, R01-NS078136, R01 NS092882-03, UH2/UH3-NS95495).