Dynamics of long range network interactions in focal epilepsy

PROJECT SUMMARY Epilepsy is the world?s most prominent serious brain disorder, affecting nearly 50 million people worldwide. For about 30% of these patients, seizures remain poorly controlled despite optimal medical management, with attendant effects on health and quality of life. In order to enable advances in the therapeutic management of epilepsy, a thorough understanding of how cellular processes that drive seizures are linked to large-scale network effects is needed. While seizures impact large brain areas and often multiple lobes, the driving processes span regions on the scale of millimeters. These have been well characterized in animal models, but the relevance to human seizures, i.e. how seizures are driven by brain signals from small-scale processes remains unclear. Instead, the view that naturally-occurring seizures may be attributable instead to large-scale neural mass effects (i.e., the epileptic network) is a subject of ongoing debate. Previously, we defined a key role for surround inhibition in shaping EEG recordings of seizures at the onset site and on small spatial scales. We now propose that surround inhibition has a dual role. On a millimeter scale, its abrupt failure permits the advance of a seizure. At long distances from the seizure focus, strong local inhibition serves to mask the excitatory effects of seizures and may help to hasten seizure termination, while weakened inhibition may permit emergence of ictal activity at a distant, noncontiguous seizure site. Multiple seizure foci may go unrecognized with standard EEG interpretation methods, and are likely a critical factor in epilepsy surgery failures. We hypothesize that once established, multiple ictal generators behave as delay-coupled oscillators, demonstrating activity that is synchronized or even temporally reversed. This results in complex and at times counterintuitive network behavior that can be challenging to reverse engineer from EEG recordings. Typically, however, even intracranial EEG recordings provide only a limited view of neural activity. In this project, an interdisciplinary research group with combined expertise in epilepsy, clinical neurophysiology, computational modeling, and mathematics will conduct a comprehensive study of the neuronal contributors to epileptic networks utilizing a unique combined dataset of simultaneous microelectrode and macroelectrode recordings of human seizures. Using a machine learning approach, we will apply this information to develop a multivariate EEG biomarker based on the inferred source of EEG discharges, high frequency oscillations, and very low frequency (DC) shifts and assess its predictive value for post-resection surgical outcome. We anticipate that the project will lead to a theoretical framework for rational development of innovative strategies for developing interventions to control seizures.