Abstracts

Rationale: Transient high frequency oscillations (HFOs) in the human electroencephalogram (EEG) in the ripple (80-200 Hz) and fast ripple (250-600 Hz) ranges have attracted increasing interest because HFOs appear to localize to seizure onset zones (SOZs), and the resection of brain tissue with high rates of HFO activity appears to be correlated with better seizure outcomes. Expert visual inspection remains the gold standard for the identification of HFOs. However, expert visual inspection is an extremely laborious process and impractical for long-term EEG data as acquired for the patient undergoing pre-surgical localization. Here we present a computer algorithm for the detection of HFOs that may help facilitate this analysis. Methods: We have developed a high resolution pseudo-wavelet time-frequency analyzer called the damped-oscillator oscillator detector (DOOD). This method uses a set of damped mathematical oscillators to detect oscillations in EEG data. The DOOD spectral density is first Z-normalized by subtracting out the mean and dividing by the standard deviation. Events with spectral density greater than a threshold S are flagged as candidate HFOs. Test data consist of 900s of 17-channel intracranial microwire EEG collected from a patient undergoing epilepsy surgery evaluation. The first 8 channels extend in a tuft from the tip of the electrode shaft. Channels 9-17 are spaced along the shaft of the electrode. Channels 1-11 are in the putative SOZ. Institutional review board authorization for this study was obtained. Results: Figure 1 shows the sensitivities (Sens) and positive predictive values (PPVs) of the DOOD HFO detection algorithm as a function of the threshold S. For comparison, sensitivities and positive predictive values using the line length method are also shown. The DOOD method is capable of higher positive predictive values than the line length method. Figure 2 shows HFO event rates using DOOD with S = 4. HFO event rates from the SOZ (channels 1-11) are higher than from non-SOZ (unpaired 2-tailed T-test, p = 0.00045). HFO event rates with a lower threshold S = 2 fail to pick out the SOZ (not shown, p = 0.31). Conclusions: DOOD may be useful for quantifying HFO event rates. A higher detection threshold S yields higher positive predictive values of HFOs and better identification of SOZ. Funding: CURE, NIH RO1-NS63039.