Abstracts

Rationale:
Epilepsy surgery is the therapy of choice for patients with medically refractory epilepsy (MRE). The prime goal of surgery is the removal of the epileptogenic zone (EZ). The gold standard for the EZ delineation is to record seizures with intracranial electroencephalography (icEEG) electrodes, which yet imply complications due to their invasiveness. Hence, the availability of interictal biomarkers that delineate the EZ non-invasively is of high importance. High-frequency oscillations below 250 Hz (ripples) are promising interictal biomarkers of epilepsy, which can be recorded both invasively, with icEEG, and non-invasively, with EEG and magnetoencephalography (MEG). Yet, the surgical value of ripples has been long debated since their generating area (ripple-zone), often includes non-epileptogenic regions. To untangle this debate, our recent study showed that ripples propagate across icEEG electrodes and the onset of this propagation (onset ripple zone) is key to estimate the EZ in children (Tamilia et al. 2018). It is yet unknown whether non-invasive techniques can map this propagation and discriminate the onset-ripple-zone from its areas of spread. The goal of this study is to assess whether high-density (HD) EEG and MEG can delineate the ripple-zone and onset-ripple-zone in children with MRE (using a virtual implantation that replicates the invasive one, Fig 1) and estimate the prognostic value of removing the virtually estimated zones for pediatric surgery.
Method:
We examined children with MRE who had surgery after long-term icEEG monitoring. We used interictal HD-EEG (72 channels) and MEG (306 sensors) to build “virtual” sensors (VSs) in the child’s brain (via beamformer techniques) at identical locations of the icEEG (Fig 1). An automated algorithm was used to detect ripples and their spatiotemporal propagation on VSs (Fig 2A) for both HD-EEG and MEG. The same algorithm was used on icEEG to define our gold standard for ripple-zone and onset-ripple-zone. Rates of ripples and onset-ripples on VSs were computed and receiver-operating characteristics (ROC) curves were built to assess their accuracy in identifying the gold standard. Finally, we assessed whether resection of the virtually defined zones predicts the individual patient’s outcome (good: Engel 1; poor: Engel≥2) via Fisher exact test.
Results:
We included 28 children (mean age: 11.5 years; 17 good-outcome). Ripple propagation was detected on VSs in all patients with higher rates on EEG-VSs than MEG-VSs (p< 0.001). EEG-VS and MEG-VS showed similar accuracy to the gold standard: EEG-VSs estimated the ripple-zone and onset-ripple-zone with an area under the ROC curve (AUC) of 0.76 and 0.70, while MEG-VS with an AUC of 0.65 and 0.64. The localization error did not differ between EEG-VSs and MEG-VSs (ripple-zone: 26-27 mm; onset-ripple-zone: 22-24 mm, p>0.05). Resection of the virtually defined onset-ripple-zone predicted good outcome with higher accuracy (EEG: 85%, p=0.005; MEG: 77%, p=0.015) than resection of the ripple-zone (EEG and MEG: 77%, p=0.016; p=0.02) (Fig 2B).
Conclusion:
Our data provide first evidence that ripple propagation can be delineated noninvasively in pediatric MRE: we showed that HD-EEG and MEG can be used to build a virtual implantation of sensors in the child’s brain, which replicates the ability of icEEG to map ripples (ripple-zone) and their propagation onset (onset-ripple-zone). We demonstrated the prognostic value of the virtually defined onset-ripple-zone in pediatric surgery, since its resection predicted the child’s outcome. The possibility to map ripple propagation via VSs possibly augments the presurgical evaluation of pediatric MRE by reducing the need for invasive monitoring.
Funding:
:RO1NS104116-01A1 and R21NS101373-01A1 by NINDS.