Abstracts

Rationale: Epilepsy surgery planning in tuberous sclerosis complex (TSC) is challenging due to presence of multiple cortical tubers and frequent occurrence of multiple seizure types. Current practice relies heavily on visual inspection of ictal and interictal intracranial video-EEG (iEEG) recordings. We previously reported interictal iEEG findings in children with epilepsy and TSC which suggested multiple, widespread high frequency oscillation (HFO) zones (Okanishi et al, Epilepsia 2014;55(10):1602-1610). The observation that resection of all distributed areas with high HFOs (>80 Hz) was associated with good seizure outcome suggests that dynamic and distributed interictal networks may contribute to epileptogenicity in TSC. We investigated the dynamics of interitcal iEEG connectivity and its relation to epileptogenic brain regions and tubers in TSC.Methods: 12 children with TSC and epilepsy, aged 3-18 years (5 females) who underwent iEEG followed by resective surgery at the Hospital for Sick Children were retrospectively studied. We analyzed 10 two-minute segments of interictal iEEG data for each patient. All possible pairwise interactions were evaluated for functional connectivity within broadband frequencies ranging from delta (1-4Hz) to ripple (150-200Hz) and fast ripple (200-300Hz) bands as defined by the phase locking value. Network structure and composition was tracked over time using eigenvector centrality (measure of importance of each electrode) to uncover dynamic connectivity maps in the interictal period. We also evaluated each connectivity map against the final resection area.Results: We found that (1) the network structure captured by eigenvector centrality is different across frequencies and even variable across interictal segments in the same patient, and (2) that network connectivity dynamically transitions through recurring connectivity maps over time. Across patients, 4-6 connectivity maps were present throughout the 20 minutes of iEEG data. With a significance level of 0.05, connectivity maps at ripple and fast ripple frequencies were most specific to capturing the seizure onset zone. In the 10 patients where MRI revealed multiple cortical tubers, connectivity map composition included engagement of multiple tubers as well as each tuber independently. Among the three patients who had poor surgical outcome (ILAE 4/5), there was a discordance between the resection area and the connectivity maps captured in the interictal iEEG data.Conclusions: The results indicate that epileptogenic networks are present and dynamically express multiple recurring network structures and compositions through the interictal period. We also show that good seizure outcome is correlated with resection of all identified connectivity maps in the HFO frequencies. The results suggests an approach to finding the epileptogenic zone without the risk of long term intracranial monitoring.