Abstracts

Rationale: In the localization of seizure onset for resective surgery, confidence increases with the number of observed seizures. Invasive intracranial monitoring may thus last for several days if seizure frequency is low. In addition, inflammation from implantation may affect characteristics of the EEG immediately afterward. Studying relatively frequent interictal epileptiform bursts over an extended period of time may provide improved localizing information to aid in surgical resection. Methods: Three dogs with spontaneously occurring epilepsy were implanted with 16 electrodes, 8 in each hemisphere, and continuously recorded for 476, 330, and 452 days respectively (Davis et al., 2011). Seizures were automatically detected, clinically validated, and onset channels identified. A high sensitivity in-house burst detector was used to identify focal bursts, followed by a culling procedure and an unsupervised Gaussian mixture model to remove false positives and separate bursts subtypes from noise. Channels of maximum average power and power in theta, alpha, beta, and gamma were calculated for each subtype and compared to clinically validated seizure onset channels. Results: Each dog had multiple seizure types, with onset channels in multiple foci. The power characteristics of each burst also evolved over time. The channels with high power in specific frequency bands (alpha, gamma) correlated with the epileptologist-determined seizure onset channels. However, there was variability between dogs. For dog 1, burst power in the gamma band was maximum in the onset channels. For dog 2, the alpha band peaked in seizure onset channels, though was less distinct from other channels. For dog 3, peaks in alpha and gamma did not clearly localize to seizure onset. Conclusions: The extended intracranial recording of dogs with spontaneous epilepsy shows an evolution of interictal burst characteristics. Power analysis of detected interictal bursts show that power in certain frequency bands are more prominent in channels that were predetermined to be located in the seizure onset area, though significant variability is present. These results suggest that the analysis of more frequent interictal bursts in humans may improve localization for the surgical treatment of epilepsy.