Abstracts

Rationale: High frequency oscillations (HFOs) are emerging biomarkers of epileptogenicity that are typically investigated with intracranial EEG (iEEG). A major limitation of iEEG is that, due to its invasiveness, it can only sample a portion of the entire cerebral cortex. Full-head non-invasive tests are often necessary to optimize iEEG placement and help tailoring the resection before surgery. To this purpose, high-density (HD) EEG and/or MEG can be recorded with the typical goal of localizing interictal spikes via electric and magnetic source imaging (ESI/MSI). Yet, beside spikes, recent evidence has showed that scalp EEG and/or MEG can also record HFOs in the ripple band (80-250 Hz). Although this offers the prospect of expanding HFOs to an earlier stage in the presurgical evaluation, the translation of scalp-recorded ripples to clinical practice is extremely limited. The goal of this study was to determine whether, in children with drug-resistant epilepsy (DRE), the localization of ripple sources (R-sources) via non-invasive ESI/MSI is concordant with the ground-truth defined by the iEEG and whether R-sources are linked to the epileptogenic zone. Methods: We included children with DRE who underwent pre-surgical HD-EEG (72-channel) and MEG (306-channel) recordings, iEEG monitoring and surgery. We detected ripples (80-160 Hz) on HD-EEG and MEG separately, and distinguished between ripples with spikes (R-wSP) or not (R-alone, Fig. 1A). Each R-source was localized using the wavelet Maximum Entropy on the Mean (wMEM) and was compared with the iEEG-defined ripple-zone (iR-zone) and surgical resection (Fig. 1B). Concordance of ESI and MSI to the iR-zone was estimated for both R-wSP and R-alone. The proportion of R-sources within the resection was compared between good (Engel 1) and poor outcome group. For RwSP, we also assessed whether R-sources and spike-sources (SP-sources) were overlapping or distinct. Results: We detected ripples on HD-EEG (n=449) and MEG (n=192) in all 16 children included in the study. Higher rates of R-wSP were found on HD-EEG than on MEG (1.1 vs 0.3 R-wSP/min; p=0.002), as also for R-alone (0.4 vs 0.1 R-alone/min; p<0.001, Fig. 2A). The overall concordance of R-sources to the iR-zone was 79% for ESI and 83% for MSI (p=0.3) regardless the overlap with spikes. MSI concordance was higher for R-wSP than R-alone (86% vs. 46%; p<0.001, Fig. 2B); while no difference was seen for ESI (80% vs. 73%; p=0.3). For R-wSP, the proportion of R-sources that were resected was associated with outcome for both ESI (good: 51%; poor: 22%; χ2: p<0.001) and MSI (good: 76; poor: 51%; χ2: p=0.003; Fig. 2C); while no association was seen for R-alone. For spikes, the resection of ESI was associated with outcome regardless the concurrence with ripples (see Fig. 2D; χ2: p<0.001); while, for MSI association was seen only for spikes with ripples (SP-wR). Finally, 52% and 43% of the sources of R-wSP were distinct from the sources of the co-occurring spikes for ESI and MSI respectively (Fig. 2E). Conclusions: Presurgical localization of scalp-recorded ripples with HD-EEG or MEG is highly concordant with the iEEG ground-truth, offering the prospect of expanding HFOs to an earlier stage in the presurgical evaluation of children with DRE. Non-invasive ESI and MSI of R-wSP, but not of R-alone, is linked to the epileptogenic zone. Furthermore, the source of ripples and spikes are often distinct, even if concurring in time, suggesting the complementarity of spike and ripple source imaging in the pre-surgical evaluation of children with DRE. Funding: R21NS101373-01A1 from the National Institute of Neurological Disorders & Stroke (PIs: Papadelis and Stufflebeam)