Automatic detection of high frequency epileptiform oscillations from intracranial EEG recordings of patients with neocortical epilepsy

Abstract

High frequency epileptiform oscillations (HFEOs) have been observed before neocortical seizures on intracranial EEG recordings. There is suggestion that HFEOs may localize epileptic brain regions important to seizure generation in humans, a finding that would be valuable for understanding, diagnosing, and treating epilepsy. In this paper, an automated approach for detecting HFEOs is described. Fuzzy clustering and histograms are used to characterize HFEO events. Compared to neurologist markings, the algorithm detected 87% of the HFEOs while achieving 68% precision and 90% specificity, without training. Applied to thirty-five minute seizure records obtained from six patients, spatial and temporal localization of HFEOs were observed in 77% and 61% of the segments respectively. Results highlight the potential of the method to identify brain regions vital to seizure generation by tracking the spatio-temporal evolution of high frequency seizure precursors in the epileptic network.