Abstracts

Rationale: Spikes are well established epilepsy biomarkers that suffer from low specificity since they can also be found in non-epileptogenic brain areas. The ability to distinguish between more and less epileptogenic spikes is highly desirable to facilitate the delineation of the epileptogenic zone (EZ) and improve their specificity. High frequency oscillations (HFOs), categorized into ripples (80-250 Hz) and fast ripples (250-500 Hz), are promising interictal biomarkers of the EZ. Fast ripples are more specific but they may not be sampled with conventional intracranial EEG (iEEG) electrodes and thus not seen in all patients. Conversely, ripples can be seen in most patients, but they have low specificity. Here, we aim to distinguish between more and less epileptogenic spikes in iEEG using the temporal and cross-frequency coupling features of spikes with HFOs.

Methods: We analyzed iEEG data from 20 children with drug resistant epilepsy who underwent resective neurosurgery. We dichotomized our patients into good (Engel 1; 10 patients) and poor outcome (Engel ≥ 2; 10 patients, ≥ 1-year follow-up). In an automated fashion, we identified spikes (S) (1-70 Hz), ripples (R), and fast ripples (FR), and sorted them into 10 groups: All S, All R, All FR, S only, R only, FR only, S+R, S+FR, S+HFOs (S+R and S+FR), and S+R+FR. For each channel, we estimated the rate of these events and computed the Mean Vector Length Modulation Index (MI) between amplitude of R and phase of S. Using receiver operating characteristic (ROC), we evaluated the performance of each biomarker to localize the clinically defined seizure onset zone (SOZ) and resection. We compared ROC curves between all biomarkers (Delong). MI values and coupling angles were compared between SOZ and non-SOZ and resected and non-resected electrodes (Mann-Whitney U, Rayleigh test).

Results: S+HFOs, S+R, and All S performed better than the other S/HFOs combinations (p < 0.001) for localizing both the SOZ and resection, but there was no difference among them (p >0.05, Fig. 1A). The MI, instead, outperformed all other biomarkers for prediction of resection (p < 0.01, Fig. 1A). MI, S+R and S+HFOs showed higher sensitivity (for specificity values ≥ 85%) (Fig. 1B,C,D). We observed a higher MI value inside (normalized mean: 0.48 ± 0.24) compared to outside resection (normalized mean: 0.39 ± 0.21) for good outcome patients (p < 0.001, Fig. 2A,B,C₁), while the same was not true for poor outcome patients (p=0.584, Fig. 2C₂). HFOs in resected electrodes exhibited strong phase locking with the slow wave (p < 0.001, Fig. 2E₁) at a mean phase angle of 338±66°. HFOs in non-resected electrodes also showed strong phase locking (p < 0.001, Fig. 2E₂) but at a mean phase angle of 41±63°.

Conclusions: Our study showed that HFOs overlapping on spikes are not better biomarkers of the EZ than spikes alone. Rather than their temporal co-occurrence, features of cross-frequency coupling, such as the MI and phase angle, between spikes and ripples can distinguish between more and less epileptogenic spikes.

Funding: Please list any funding that was received in support of this abstract.: RO1NS104116-01A1 and R21NS101373-01A1 by NINDS.