Abstracts

 Methods: 256,159 ECoG recordings (ECoGs) were collected at scheduled times of day (usually once per day) from 206 subjects over a median 6.8 years. Each record was typically 90 seconds of 4-channel ECoG data recorded at a sampling rate of 250 Hz. Interictal spike rate, total power, and power in different frequency bands were computed for each ECoG channel (EC). Correlations between these ECoG features and subject-reported clinical seizure rates (CSRs) were measured. Three complimentary methods were used to compute correlations between CSRs and the selected ECoG features. In the 1st method, correlation coefficients (CCs) were computed between ECoG features and CSRs averaged in 28-day overlapping windows with a window increment of 1 day. In the 2nd method, to avoid oversampling of data points, CCs were computed between ECoG features and CSRs averaged in 28-day non-overlapping windows. In the 3rd method, to give every ECoG equal weight, CCs were computed between features from individual EC and averaged CSRs using a 28-day window spanning 14 days around 00:00 on the day of each ECoG. In all 3 methods, each EC was analyzed individually and a minimum of 10 data points was required to qualify for computing the CC. Results: Significant correlations (SCs; p < 0.05) was found between ECoG features and CSRs in a subset of ECs using CCs computed by all 3 methods. Using the 1st method (n = 792), 77.2% of ECs had SCs between spike rate and CSR. Of these significant ECs, 68.7% had positive and 31.3% had negative CCs. Within the significant ECs, absolute CC values were >0.25 for 33.9%, and >0.50 for 13.5% of ECs. In patients with mesiotemporal seizure onsets, 81% of ECs showed SCs between spike rate and CSR (71.8% positive, 28.2% negative). Among the patients with neocortical seizure onsets, 81% of ECs had significant correlations (64.1% positive, 35.9% negative). SCs were also observed between the other ECoG features and CSRs. Conclusions: Analysis of scheduled, stable, ambulatory ECoG data from 206 patients from the previously completed RNS System clinical trials and continuing long-term follow-up study has shown that certain ECoG features such as interictal spike rate are significantly correlated with subject-reported clinical seizures and hence may be regarded as seizure biomarkers in some patients. With further investigation, biomarkers obtained from long-term ECoG recordings could be used to rapidly assess the efficacy of antiseizure therapies. Funding: None.