Abstracts

Rationale: High frequency oscillations (HFOs) are becoming a well established biomarker of epileptic tissue. However, translation of this biomarker to clinical settings requires overcoming both scientific and technical challenges. One technical concern is the unknown required and optimal sampling rate and anti-aliasing filter position used when recording HFOs with intra-cranial EEG. Manufacturing companies are encouraging hospitals and clinics to replace their EEG recording equipment with newer models featuring faster sampling rates, but it is not yet known how high of a sampling rate is actually clinically useful.Methods: This study addresses that need using deidentified data acquired from 17 patients undergoing intra-operative monitoring for epilepsy surgery. The intra-cranial EEG data were acquired at greater than 20 kHz, then initially down-sampled to 5 kHz with an anti-aliasing filter position at 2 kHz. A standard set of HFO and artifact detections were determined from the data. HFOs were additionally detected with the data down-sampled to lower sampling rates (500 Hz to 2.5 kHz) or with an additional anti-aliasing filter applied (filter positions from 100 Hz to 1.5 kHz). HFO detections coincident with detected artifacts were redacted. The number of HFOs detected for each sampling rate and/or filter position were compared with the standard set of HFOs detected at 5 kHz--over 1 million HFOs.Results: Many results were as expected, with the number of detected HFOs at or above 90% for anti-aliasing filter settings above 500 Hz, and a steady drop in detected HFOs with respect to sampling rate—already dropping to about 80% at 2.5 kHz. Despite the dramatic changes in HFO detection sensitivity, moderate changes were observed in the HFO feature distributions and only minor changes were observed regarding the clinical correlation of HFOs with epileptogenic tissue.Conclusions: Sampling rate and acquisition filter settings have large effects on the number of HFOs detected. However, even when low resolution results in significantly fewer HFOs, those that are detected are still potential biomarkers of epileptic tissue.