Harmonic patterns embedding ictal EEG signals in focal epilepsy: a new insight into the epileptogenic zone

Abstract

AbstractThe ictal EEG biomarkers of the epileptogenic zone (EZ) require more precise definition. The ictal initial fast activity is important in EZ localization, but EEG onset patterns are heterogeneous and the fast activity is absent in many patients. Here we defined a unique spectral structure of “harmonic pattern” (H pattern) on stereo-EEG (SEEG), characterized by multiple equidistant, high-density bands with varying frequency on time-frequency map. The H pattern was commonly observed among 57 (81.4%) out of 70 patients with focal onset pattern on SEEG. It manifests in seizures with various ictal onset patterns with or without fast activities, occurring at either early or late stages of seizures. The H pattern was found to be embedded only in two types of EEG segments: fast activity with a frequency > 25 Hz (FA-H pattern) during early stage of seizure (mean 13.3 seconds after onset), and irregular polyspikes (> 5 Hz, PS-H pattern) observed during late stage (mean 23.3 seconds after onset). The H pattern is typically observed at a very proximate time point across the seizure onset zone (SOZ), the early propagation zone, and occasionally other areas, with the same fundamental frequency, indicating an inter-regional synchronization. Notably, the SOZ exhibited the highest proportion of channels expressing H pattern, and also the highest band number of H pattern. At the patient level, the dominant H pattern was defined as those with high ranks in band numbers, specifically exceeding the third quartile (Q3). Resection of the region expressing dominant H pattern, but not SOZ, independently predicted seizure freedom after surgery. This suggests that the dominant H pattern serves as an ictal marker of EZ. The generation mechanism of the H pattern was then explored. Through nonlinear analysis, our data demonstrated that the H pattern was generated by specific nonlinear phenomena rather than being a result of frequency intermodulation or purely methodological artefact. The degree of nonlinearity was more pronounced in the dominant H pattern compared to non-dominant counterpart. We postulate that FA-H pattern may be sustained by a predominant and synchronized firing of GABAergic neurons, while excitatory neuron firing plays a more important role for PS-H pattern. As a distinctive and common ictal spectral feature, H pattern conveys unique information of ictal neural dynamics and provides new insights into the EZ. Our study additionally presents evidence supporting the existence of an elongated time-window for measuring EZ through quantitative EEG.