Abstracts

Though highly effective and safe for treating refractory focal epilepsy, resective and ablative epilepsy surgeries are not available to many people with epilepsy--for example, if the seizure onset zone (SOZ) is in an essential area of the brain or if there were multiple SOZs. For these individuals, responsive neurostimulation (RNS) offers an additional treatment with minimal risks of causing permanent neurologic deficits. Candidates for RNS typically need to undergo intracranial EEG monitoring (phase 2) to locate the SOZs, which can be difficult due to high variability in epilepsy etiology, surgical approach, electrode location, and ictal onset patterns, and differences in EEG acquisition and interpretation can contribute to inconclusive results and localization errors. Ictal high frequency oscillations (HFOs) are a reliable marker of SOZs, and our work seeks to study ictal HFOs a simple, generalizable feature for targeting RNS.

With IRB approval, we retrospectively reviewed the records of adult patients at UC Irvine from 2015 to present. Our cohort included all who underwent both phase 2 and RNS implantation (with depth or cortical electrodes, or both) and had at least 2 years of follow-up, including those with prior epilepsy surgeries, but excluding those who underwent combined resection/ablation plus RNS. While blinded to outcome data, we analyzed ictal EEG (at least 3 seizures of each SOZ if available) with an assistive custom Matlab algorithm that quantifies and visualizes the spectral power of different frequency bands across time (focusing on HFOs). EEG data were sampled at either 500Hz or 1000Hz, and both depths and subdural electrodes were used, sometimes in combination. Ictal HFOs were seen with both electrode types and in both mesial temporal and cortical regions. Finally, we reviewed pre- and post-implant seizure reporting to determine the Engel class, and then examined the association between RNS targeting ictal HFOs and outcomes.

A total of 26 patients met the inclusion criteria. Of these, 21 patients (~81%) had significant improvement of seizure control (7 Engel I, 3 Engel II, 11 Engel III) at 2 years postimplant, not factoring in any medication changes. Of the 5 with poor outcomes, 2 patients had infected devices within 2 years of implantation and were explanted, and 1 patient died suddenly, unexpectedly within 1 year. None had additional epilepsy surgery within the 2-year period. Of the 23 patients with intact RNS at 2 years, all of the 7 Engel I patients had RNS targeting SOZs with ictal HFOs (including HFOs + spikes/rhythmic). For the other 16 with Engel II-IV, 6 had RNS targeting HFOs and 10 did not. The Fisher’s exact test p-value was 0.0075 for this 2x2 contingency table.

Ictal HFOs are a reliable indication of SOZs, and precise targeting of these SOZs with RNS electrodes seems to lead to better seizure control. HFOs can be easily missed, and it is important to have the proper setup (preferably 1000+Hz sampling) and settings for analysis (filter and gain settings). More sophisticated computational methods may improve the specificity further.