Abstracts

Rationale: Epileptogenic networks are thought to be closely associated with seizure generation and propagation, which provide a new perspective for epilepsy diagnosis and therapy. Current network measures mostly focus on exploring the electrophysiological oscillations within specific frequency while omitting interactions of different frequencies. Some studies suggested that low-frequency (LF) to high-frequency (HF) coupling during the ictal period is a useful biomarker for localizing epileptogenic tissues. However, it is still unclear whether the LF hubs or HF hubs of the cross-frequency coupled network (CFCN) is more critical for treatment outcome. We hypothesis that HF hubs of the CFCN are primarily responsible for seizure dynamics other than LF hubs, which are fundamental in planning for surgery. Methods: Stereo-electroencephalography (SEEG) recordings of 36 seizures from 16 patients with intractable epilepsy were analyzed. All the patients had undergone presurgical evaluation and surgical or radiofrequency thermocoagulation treatment. Informed consent was obtained from each patient and the study was approved by the ethics committee/IRB of the affiliated institutions. Here, we proposed a novel approach for modeling the CFCN in epilepsy patients. First, coupling strength across paired channels were calculated in the form of four-dimensional modulation matrices. By treating channel domain and frequency domain independently, principle component analysis (PCA) was applied to the channel domain of the matrix. Next, we extracted the first principle component in the frequency domain as the principle modulation index, and the corresponding PCA contribution coefficients, as the CFCN. Graph measures were then applied on the CFCN to characterize HF hubs and LF hubs. The distributions of those hubs were further evaluated in relation to recording sites within and outside the regions of treatment. Results: The amplitudes of ictal high-frequency oscillations (HFOs) (30–150 Hz) were coupled with the phase of low-frequency oscillations (LFOs) (1–13 Hz) dynamically over time, frequency bands, and space. In Engel Class I group (n=10), the strength of HF hubs was significantly higher inside the treated regions compared to other regions during all stages of seizure. In Engel Class II group (n=4), the strength of HF hubs was only significantly higher compared to non-treated regions only at the early stage of seizure. There was no significant difference found in Engel Class III group (n=2). With the CFCN calculated for the whole seizure duration, 92.5%, 33.3%, and 0% of the “active” HF hubs were treated in patients with Engel Class I, II, and III outcome, respectively. The strength of HF hubs inside the treated regions was significantly higher than those outside the treated regions in Engel Class I group (p<0.001). However, the LF hubs did not show correlation with outcome. Furthermore, our SEEG analysis showed that deep epileptogenic sources also exhibit the cross-frequency coupling features. Conclusions: Our findings suggest that HF hubs but not LF hubs of the CFCN predict treatment outcomes, which lead to improved understanding of the coupling between the HFOs and LFOs in seizure dynamics. The proposed method allows us to localize pathological HF hubs in the epileptogenic network with high precision, and potentially to achieve more targeted surgical interventions or neuromodulation therapies. Funding: CL is supported in part by grant from the Natural Science Foundation of China (61771323).