Abstracts

Rationale: Recent studies have demonstrated the utility of pathological high frequency oscillations (pHFOs) as an important biomarker in localizing seizure onset zones (SOZ). With modern implantable neuromodulation devices now capable of longitudinal recording, characterizing pHFO patterns prior to seizure onset may improve algorithms for early seizure detection and perhaps even seizure prediction. Prior studies have reported on the inconsistent temporal behavior of pHFOs among patients near the seizure onset time. Here we demonstrate that using smaller time windows for analysis can yield more consistent pHFO patterns.

Methods: Data were obtained from patients undergoing presurgical intracranial EEG recordings for medically refractory epilepsy at Stanford Health Care. Our study was approved by the Stanford Institutional Review Board, and informed consent was obtained from all patients. To limit sampling bias due to electrode placement, we included only patients with clear lesional focal epilepsy and congruent presurgical diagnostic data regarding the location of seizure onset (i.e., clinical semiology, imaging, neuropsychological testing, and scalp EEG). Intracranial electrodes were implanted to confirm the SOZ and identify the extent of areas to be resected or targeted for neuromodulation. EEG data were recorded at 1000 samples per second, using a combination of depth, grid, and strip electrodes as clinically indicated. EEG segments at regular time intervals between seizures (interictal) and at specified intervals just prior to seizures (preictal) were selected for analysis. Using an in-house automated detection, electrographic events were identified and sorted based on time and frequency features. To minimize physiologic high-frequency activity, analysis was limited to electrodes near the seizure onset zone and symptomatic regions, as determined by the attending clinicians.

Results: Interictal and preictal pHFO patterns varied among patients, but for a group of patients there seemed to be a common behavior of a transient drop in the rate of preictal (and at times the rate of epileptiform discharges as well). The exact timing and degree of the change varied among patients and to a lesser extent among individual seizures within each patient, but some patients had a marked drop in HFO rate, seen only when using a smaller time window on the scale of 1 minute. In the preictal EEG segments, there may be an overall increase in pHFOs or otherwise equivocal change, so this phenomenon would be missed with longer time windows. The fluctuations of pHFOs in the interictal EEG were not as pronounced.

Conclusions: The results are preliminary, and the mechanism underlying the pHFO pattern is unclear, but finding patterns may improve methods for individualized early seizure detection and prediction, which could not only provide a better warning system and lead to more effective treatment.

Funding: None