Abstracts

Rationale: The gold-standard biomarkers for estimation of the epileptogenic zone include the seizure-onset zone (SOZ) involved in generation of habitual seizures. SOZ is associated with an increased rate of interictal spike discharges [1,2] as well as a tight coupling between interictal high-frequency oscillations (HFOs) and slow-wave at 3-4 Hz [3,4]. A recent study reported on a computational model for structural connectivity-based simulations from diffusion tensor imaging (DTI) to predict the likely locations of seizure onset zone(s) [5]. The present study explored whether a "DTI-connectome based seizure initiation model" can be an effective non-invasive biomarker to estimate the epileptogenic zone in pediatric epilepsy surgery. Methods: Pre-operative DTI, 3D-T1 freesurfer cortical thickness data and subdural EEG monitoring were acquired from 12 children with left temporal lobe neocortical epilepsy (age: 9.25.2 y.o., 6 boys) with (n=8) or without (n=4) extratemporal involvement. A total of 1015 cortical regions of interest (ROI) were parcellated to construct a 1015x1015 connectivity matrix, S in which the element Sij quantifies the pair-wise connectivity score (i.e., fiber numbers connecting any two given cortical regions normalized by total volume of two ROIs, 1 <= i,j <=1015). At the ith ROI, betweeness centrality measuring the number of neighboring axonal paths that passed though and average cortical thickness were assessed to model the likelihood of seizure in the bistable seizure initiation model [4] where temporal seizure activation level at time t, zi(t) was numerically determined by a complex non-linear function of four parameters including inter-regional connectivity score, betweeness centrality, cortical thickness and inter-regional delay time. The seizure escape time of the ith ROI measuring how long it takes for the ith ROI to transit to the seizure state was estimated by the time of t satisfying the real value of zi(t) > 1 (i.e., a shorter escape time related with higher epileptogenicity resulting from altered connectivity and cortical thickness). Finally, the subtraction of HFO coupled slow-wave at frequency of 3-4Hz from 0.5-1Hz, |MI(HFO)&(3-4HZ)-MI(HFO)&(0.5-1HZ)| was measured using subdural EEG at SOZ and then compared with the corresponding seizure escape time using support vector machine analysis. Results: An apparent increase in betweeness centrality and cortical thickness was found at the ROIs that were identified as SOZ, associated with shorter seizure escape times in the electrodes of SOZ as compared to electrode showing only interictal spiking or non-epileptogenic electrodes (Fig. 1 and 2). Support vector machine analysis found that the value of seizure escape time provides an accuracy of 0.76 to predict the SOZ (sensitivity/specificity=0.77/0.76). In the same analysis using the value of |MI(HFO)&(3-4HZ)-MI(HFO)&(0.5-1HZ)|, a similar accuracy of 0.77 was achieved (sensitivity/specificity=0.88/0.66). Conclusions: The present study suggests that the non-invasive DTI-connectome based seizure initiation model may provide a stand-alone tool to predict the SOZ in pediatric epilepsy surgery. Funding: This study was funded by a grant from National Institute of Neurological Disorders and Stroke (R01-NS089659 to J.J).