Abstracts

Rationale: HFOs have been defined as a novel biomarker for the epileptogenic zone that reflects the underlying tissue's pathological power. Finding HFOs could better delineate resection areas in patients with refractory epilepsy undergoing surgery and improve epilepsy surgery outcomes. Identifying HFOs is a challenging task, and the exact clinical definition of their features is still unclear. HFOs can be divided into normal HFOs (nHFO) and pathological HFOs (pHFO), and each has different signal characteristics. In this work, we analyzed the resting-state ECoG signal to differentiate nHFO from pHFO by using the potential of the IEDs biomarker.

Methods: Five patients underwent ECoG electrode implantation at the Megumino Hospital in Japan. Data were recorded in the epilepsy monitoring unit (EMU) at night during sleep resting-state. ECoG signals were reviewed to find seizure onset zone (SOZ) and confirmed with video EEG recorded in EMU. An automatic event detector identified HFOs and IEDs, and then time points of the evets were compared to find whether they overlap together. If there is an overlap between IEDs and HFOs, the event is considered pHFO, and HFOs without IEDs are considered nHFO. Briefly, the detector algorithm filters signal (5-60 Hz for IEDs and 60-250 Hz for HFOs detection), apply Hilbert transform to calculate signal envelop as energy function, and epochs it into minute-long length. Then, robust z-scoring is applied to normalize the envelope, and events greater than 3-SDs above the median value are considered events of interest (EOIs). Finally, detection criteria (duration >6 ms, minimum distance between two consecutive EOIs >50 ms, amplitude >5 uV, and oscillations number >4) were set to fulfill HFOs and IEDs detection. After HFO detection, the positive predictive value (PPV) was calculated for finding the relationship between the HFOs and SOZ.

Results: Using automatic HFOs detection, 120 HFOs per electrode on average (range 0-40/min) were identified from 689 electrodes in five patients with 157.3 hours of data in total. Thus, 12.8% of the HFOs were identified and average HFO rates were nHFO(SOZ)=5/min, nHFO(non-SOZ)=3.28/min, pHFO(SOZ)=3.72/min, and pHFO(non-SOZ)=0.28/min. The SOZ could be determined with PPV(nHFO)=0.6, and PPV(pHFO)=0.92.

Conclusions: From a clinical perspective, pathological and physiological HFOs could exist in all brain structures and occur either in SOZ or healthy areas. Preliminary results show that pHFO, identified by overlapping IEDs and HFOs, could predict SOZ, whereas nHFO is not SOZ specific.

Funding: Not applicable