Abstracts

Rationale: Intracranial EEG (iEEG) is often performed to define the epileptogenic zone (EZ) before epilepsy surgery. Given the unpredictability of seizures, interictal iEEG biomarkers of the EZ are of paramount importance. High-Frequency Oscillations (HFOs) are promising EZ estimators; yet, quantifying HFO rates across iEEG contacts is of uncertain presurgical value at the individual patient level. Since conventional HFO rates are ‘local’ measures that disregard the interconnections between brain regions, we hypothesize that the EZ anatomy can be better-defined by localizing the HFO-generators that are also functionally connected to other HFO-generators, creating a hyper-connected HFO-generating network.

To test the hypothesis, we proposed the “Interconnected-HFO-Index” (IC-HFO-Index) as a novel method for quantifying HFOs on brief iEEG, and assessed whether this improves upon ‘conventional HFO rates’ in predicting surgical outcomes of children with drug-resistant epilepsy (DRE).


Methods: We studied interictal iEEG of 60 children with known postsurgical Engel outcome (36 good outcome; mean 112 contacts/patient). For each iEEG contact, we detected ripples (R: 80-250 Hz) and fast-ripples (FR: 250-500 Hz) (Fig 1A-B) and computed their ‘local’ rates (per min). We measured functional connectivity (FC) between contacts (Fig 1C) in the HFO range (R, FR) but also in the slow frequency range (3-4 Hz) since pathological HFOs are coupled to these slow rhythms. We computed the IC-HFO-index of each contact by summing the contact’s HFO rate with the rates of its functionally-connected contacts (Fig 1D). We identified the topmost “interconnected HFO generators” through patient-specific thresholds (Fig 1E), measured their resection overlap and compared it between good and poor outcomes (Wilcoxon rank-sum).

Finally, we tested whether resection overlap predicts patient’s outcome and estimated ROC AUC, sensitivity, and specificity. Same analysis was done for conventional HFO rates and for clinically-defined seizure onset zone (SOZ) to test their ability to predict outcome compared to the proposed IC-HFO-index.


Results: The overlap of the interconnected HFO generators with resection was higher in good than poor outcomes for both R and FR (Fig 2 A&B; p< 0.05; E-score: R=3.4, FR=3.5). Using conventional HFO rates, resection overlap did not differ between outcome groups (R: p=0.30, E-score=1.7; FR: p=0.57, E-score=1.7) and the outcome prediction at the individual patient level resulted in a low AUC (0.54-0.58, Fig 2 in blue) suggesting minimal prognostic value.

Resection of the interconnected HFO generators instead showed high predictive value (AUC=0.75 for FR and 0.73 for R), outperforming also SOZ resection, which was not predictive of outcome (Fig 2 C).


Conclusions: The “Interconnected-HFO-Index” is a novel interictal EEG biomarker that allows to deconstruct the “HFO-generating network”, as opposed to the conventional “HFO zone”. Our findings show that localizing the interconnected HFO generators facilitates EZ localization in children with DRE, as their removal provides high prognostic value, contrary to traditional HFO or SOZ estimates.


Funding: R03NS127044 by the NINDS of NIH