Abstracts

Rationale: Identifying reliable features that occur before seizure onset is essential in developing new therapies and seizure prediction devices. Despite extensive research in this field, a consistent paradigm for seizure prediction has not yet been found. These methods have almost exclusively relied on the electroencephalogram (EEG).Methods: In the present study we used a unique microelectrode array system to investigate preictal neuronal activity. These arrays record field potential, which reflects synaptic activity of small groups of neurons, and simultaneously measure multi-unit activity, reflecting population action potentials. This approach provides cortical layer-specific information about excitatory and inhibitory dynamics and single cell firing characteristics in vivo at a resolution beyond that of conventional EEG. We performed simultaneous long-term EEG and microelectrode recordings in 8 patients with intractable focal epilepsy (5 males, ages 10-34) and captured 14 seizures. Arrays were inserted near the seizure focus in the temporal or frontal cortex for all subjects. We performed spectral (5-300 Hz) and broadband power analyses of multiunit activity during a preictal period of 30 minutes prior to the first observed EEG change. We analyzed data from an average of all electrode channels and then divided it into three groups of channels, roughly corresponding to the upper, middle and lower cortical layers, characterized by different input/output characteristics. We defined increased activity in the preictal period as a significant (> 2 S.D.) change from the average baseline taken from the first 100 seconds of the preictal period. In order to get a sense of the sensitivity and specificity of this technique, we performed these same analyses on 30 min. blocks of “interictal” data, defined as data more than 30 minutes before or after a seizure.Results: Analysis of the MUA averaged across all cortical layers revealed a significant preictal increase in 8 of the 14 seizures. Increase MUA was seen from 200-1700s before the seizure (mean ±S.D., 1163 ±446s). Of all channels, sensitivity and specificity were found to be highest for those corresponding to the lowest cortical layer: these values were 86% and 69%, respectively. In contrast, data from an average of all cortical layers yielded a sensitivity of 57% and a specificity of 71%. Sensitivity and specificity of the data from the upper and middle channels were lower. Of the subjects showing a positive preictal increase, all but one were implanted within the seizure focus.Conclusions: These preliminary results suggest that multiunit activity may increase before the seizure is apparent on EEG. These changes were seen most often in channels thought to correspond to lower cortical layers, which implies that deep cortical layers may figure prominently in the pathogenesis of ictal events. Thus implanted microelectrodes, which are capable of recording an information-rich signal as compared to typical EEG, offer a novel approach to seizure detection, prediction, and localization.