Abstracts

Rationale: High-frequency oscillations (HFOs) are electrophysiological events in the brain that exhibit distinctive temporal and frequency patterns. The distributions of HFO rates and amplitudes have been recognized as candidate biomarkers for localizing the seizure onset zone (SOZ) in patients with refractory epilepsy [1]. However, these biomarkers are sensitive to the accuracy of HFO detection methods at a single-channel level, which is low for some subjects [2]. Recent evidence suggests that measuring connectivity between high-frequency (HF) events can improve localization [3]. Thus, we developed the channel-level connectivity dispersion (CLCD) metric to quantify the variability in synchronization between high-frequency bursts in the 80-200Hz frequency band. This metric allows us to identify clusters of electrodes with abnormal synchronization, which we hypothesize to be associated with the SOZ.

Methods: We retrospectively analyzed the intracranial EEG of 89 patients from an open-access dataset [3], selecting three non-overlapped two-minute epochs per subject. We identified high-frequency bursts within each channel by filtering with a novel Gabor-based filter that maintains HFO morphology and attenuates wideband artifacts. Then, we binarize the instantaneous amplitudes of the filtered signals. From these binary sequences, we estimated covariance matrices (CMs) at frequencies ranging from 80Hz to 200Hz, combined into a single cross-frequency CM. Then, the CLCD is calculated by measuring the variability in each electrode's combined connectivity.

Results: CM clusters independently distinguished SOZ from non-SOZ channels in each subject. Total CLCD also demonstrated significant separability performance for identifying SOZs with average values for absolute difference of 0.85, ROC-AUC of 0.73, and Cohen's d of 1.05. On average, SOZ channels had lower CLCD values than non-SOZ channels. This suggests that CLCD could significantly assist in identifying SOZ clusters and, therefore, surgical planning for epilepsy patients.

Conclusions: Connectivity in high-frequency bursts, through the CLCD, seems to localize the SOZ and exhibit consistent intra-frequency and subject-level patterns. CLCD, therefore, has the potential to be a tool for surgical planning. However, further analysis is required, as validation based on the resected volume and the patient's surgical outcome is critical.

Funding: NIH NINDS R01NS116273.