Abstracts

Rationale: Whether seizures are the result of networks or focal phenomena is debated. Previous human and animal studies have provided evidence for the latter in neocortex, where two seizure territories with contrasting neural firing patterns thought to represent actively seizing vs unrecruited tissue was demonstrated. We hypothesized that ictal neural activity markers across scales would segregate similarly in limbic structures with equivalent dynamics.

Methods: 19 patients undergoing sEEG monitoring consented to inclusion of Behnke-Fried microwire electrodes. 5 measures utilizing local neuronal firing along with high gamma activity (HG, 80-150 Hz) from the nearest macroelectrode were collected over an epoch from 5 minutes preictal until transition to an ictal pre-termination pattern or clinical secondary generalization. Single units were tracked through the ictal transition using template-matching. Principal component (PC) analysis with k-means clustering assessed for structure within the data. Results were compared to clinically defined seizure onset and spread areas—the “epileptic zone” (EZ). A 10-minute interictal sleep epoch was chosen for automatic high frequency oscillation (HFO) detection in the 80–500 Hz band. Putative cell-types were determined via mean waveshapes and firing patterns.

Results: 52 seizure recordings met quality criteria. The variance ratio criterion identified two clusters as the optimal solution in PC space with 19/20 EZ sites and higher HFO rates correlating to one cluster (p < 0.01 two-tailed t-test). HG phase-locking to the dominant ictal rhythm (PLV) was the largest contributor to the top 3 principal components, explaining 45% of the variance in the data vs 3-21% for other metrics. All single metric comparisons demonstrated cluster differentiation (Bonferroni-Holm corrected rank-sum; p < 0.05). On a single unit level, there was a significantly greater reduction in putative inhibitory neuron firing rate in the EZ vs non-EZ group (p < 0.05; rank-sum). 8 units (6%) across both groups showed a cessation of firing prior to termination, and only 1 showed a significant reduction in amplitude prior to quiescence (p < 0.01; Bonferroni-Holm corrected rank-sum).