Abstracts

Rationale: The normal brain is increasingly seen as a dynamic system dependent on the integrity of structural and functional networks. Previous animal and human studies have shown that disruption of these networks may lead to epilepsy. Nodes of these networks may help identify optimal targets for surgical resection. Here, we assess the ability of functional connectivity (FC) measures to quantify the epileptogenic status of a brain area at a specific time course and predict the surgical outcome in children with drug resistant epilepsy (DRE) with intracranial EEG (iEEG). We hypothesize that FC measures can discriminate between different epileptiform states, identity the epileptogenic zone, and predict surgical outcome.

Methods: We analyzed iEEG data recorded from 31 children (11.41 years ± 5.94) with DRE who underwent epilepsy surgery. Patients were dichotomized in having good (Engel I, 21 patients) or poor (Engel II-V, 10 patients) outcomes. Contacts of iEEG recordings were classified as inside or outside the seizure onset zone and resection (Figure 1A). On iEEG data, we identified segments of one-minute duration from the following states: (1) interictal activity with no spikes; (2) interictal activity with spikes; (3) pre-ictal; (4) ictal; and (5) post-ictal (Figure 1B). For each patient, we computed and averaged the Amplitude Envelope Correlation (AEC) in segments of 3 s duration in each state for physiologically relevant frequency bands (i.e., delta, theta, alpha, beta, low-gamma, and high-gamma) (Figure 1C). We considered iEEG contacts as nodes and AEC as edges and further computed the nodal strength as the median AEC of all edges connected to itself (Figure 1D). Nodal strength was used to predict the epileptogenic nodes within each network in the Receiver Operating Characteristic (ROC) curves. We finally compared nodal strength between states (Wilcoxon signed-rank test).

Results: Nodal strength was different among states for several frequency bands (Figure 1E): it was higher during ictal and post-ictal states compared to interictal and pre-ictal states in most frequency bands (Figure 1E) (p< 0.05; FDR corrected). In patients with good outcome, nodal strength was higher inside (compared to outside) resection for all states in the beta band, and for all states (except the pre-ictal state) for the low- and high-gamma bands (Fig. 2) (p< 0.05; FDR corrected). No differences among states were observed for patients with poor outcome for any frequency band (p >0.05). The area under the curve (AUC) from ROC analysis revealed that both interictal with no-spikes (AUC=0.61) and with spikes (AUC=0.66), as well as pre-ictal (AUC=0.66) states were predictive of the epileptogenic nodes.

Conclusions: FC measures can discriminate epileptogenic states and predict surgical outcome in children with DRE. Such measures are promising biomarkers for seizure prediction and identification of the epileptogenic zone augmenting the presurgical evaluation of patients with DRE.

Funding: RO1NS104116-01A1 and R21Ns101373-01A1 by NINDS