Abstracts

High-frequency oscillations (HFOs) in neural signals hold great potential as biomarkers for various neurological conditions, including epilepsy and neurodegenerative disorders. However, accurately detecting HFOs in noisy electrophysiological data remains a challenging task.

This abstract presents a cutting-edge method for detecting HFOs with exceptional specificity, enhancing our ability to study these oscillations in clinical and research settings.

Our previous study (Cho et al., bioRxiv 2023) proposed a neural oscillation detection method called Cyclic Homogeneous Oscillation (CHO), which leverages physiologically motivated signal processing techniques to achieve superior HFO detection performance. CHO integrates 1/f noise removal, time-frequency analysis, and auto-correlation analysis to reject spurious harmonic oscillations and to identify and characterize neural oscillations within complex electroencephalography (EEG), electrocorticography (ECoG), and stereo-EEG (SEEG) signals.

We have shown the high specificity of CHO in detecting neural oscillations using synthetic and empirical data (EEG, ECoG, and SEEG signals in a total of 27 subjects).

Here, we applied the CHO method to determine whether it can detect HFOs and identify their durations, fundamental frequencies, and onset times using SEEG signals recorded in six patients with temporal lobe epilepsy.

CHO successfully detected the patient-specific fundamental frequency of HFOs and identified their locations, durations, and onset times, as shown in Figure 1. Figure 1A shows the spatial trajectory of detected 83 Hz HFOs. The onset of the detected HFOs was validated against seizure onsets determined by an experienced epileptologist. Figure 1B shows the median duration of HFOs at the time of seizure zone was 50 ms. Furthermore, the onset times of HFOs within the seizure onset zone detected by CHO were more specific than those of the Epileptogenicity Index (EI, Bartolomei et al., 2003), as shown in Figure 1C.

The concordance of EI and CHO with seizure onsets, as determined by an experienced epileptologist over 26 clinical seizures, shows that CHO is equivalent to EI. However, while EI uses theta, alpha, beta, and gamma frequency bands to detect seizure onset, CHO detects HFOs based on their onset time within a single frequency band (60-200 Hz).  Thus, combining different frequency bands within CHO may develop into a new epileptogenicity index method.

In conclusion, CHO presents a significant methodological advancement in the tools available for HFO detection. CHO demonstrates high specificity in detecting HFOs, even in noisy and challenging environments, making it a valuable tool for clinical diagnosis and basic neuroscience research.