Estimating the dominant frequency of High Frequency Oscillations in depth-EEG signals

Pathological high-frequency oscillations (HFOs, 200–600 Hz) observed in depth-EEG and on scalp EEG recordings are recognized to be potentially valuable biomarkers of the epileptogenic zone responsible for generating seizures. Many research studies have been dedicated to detect, classify, simulate and understand the underlying mechanisms responsible for their generation. However, broadly classifying the HFOs into classes of wide frequency bands may negatively impact the quality of information carried by these electrophysiological biomarkers. In this paper, we perform a comparative study of various signal processing methods for estimating the dominant frequency of HFOs. The novelty is to make use of a physiologically-plausible computational model in which the HFO frequency can be tuned a priori. Results indicate that non-parametric methods best estimate the frequency of the low-amplitude fast oscillations characteristic of HFOs.