Abstracts

Rationale: The mesial temporal lobes are highly epileptogenic, yet are often poorly recorded on scalp EEG. 25% of patients with temporal lobe epilepsy (TLE) have seizures recorded on intracranial electrodes that show no seizure correlate on scalp EEG, here termed "no-scalp seizures" (Ebersole, Epilepsia 1996). Yet, while intracranial electrodes detect deep seizures with high sensitivity, the risks of such invasive monitoring are seldom warranted. The ability to non-invasively detect deep temporal lobe seizures using scalp EEG alone would be of high clinical utility. As a first step towards this goal, we analyzed scalp EEG during "no-scalp seizures" to determine whether changes in spectral features occur during these seizures, despite the lack of a scalp ictal correlate apparent to electroencephalographers. Methods: Data was collected from patients with refractory TLE undergoing presurgical evaluation with simultaneous recordings from bilateral foramen ovale and scalp electrodes (10-20 system). We used clinical EEG reports to identify 50 "no-scalp seizures" from 13 patients. Spectrograms were generated and power was averaged within the delta, theta, alpha, beta, and gamma bands, and within the ipsilateral and contralateral hemispheres (relative to seizure onset) and scalp vertex, yielding 15 frequency-region combinations. For each "no-scalp seizure," we sampled 5000 random interictal segments equal in length to and occurring within 3 hours of the seizure. We compared each interictal segment to a preceding 2 minute baseline. Differences in median power between each interictal segment and its baseline were used to generate a population distribution for each frequency-region combination. Differences in median power between each "no-scalp seizure" and its baseline were then compared to the population distributions. Values > 97.5th or < 2.5th percentile of the population distribution were deemed statistically significant. Results: 90% of "no-scalp seizures" demonstrated a significant change in spectral content in at least one frequency-region combination. 76% of "no-scalp seizures" showed an increase in power in at least one frequency-region combination, whereas 38% showed a decrease in power. Seizures with significant increases in power in at least one frequency-region combination showed, on average, increases in 6 frequency-region combinations, suggesting that these changes are robust. Increases in power were most common in the ipsilateral hemisphere, and in the theta, alpha, and beta frequencies. Conclusions: Deep temporal lobe seizures that show no ictal correlate on scalp EEG nevertheless show significant changes in spectral content on scalp EEG. These changes may reflect volume conduction of ictal activity from the mesial temporal lobes, or cortical alterations occurring simultaneously with mesial temporal seizure onset. In either case, these changes may be pertinent for identifying deep seizures at the scalp, without the need for more invasive monitoring.