Guiding epilepsy surgery using network models and Stereo EEG

More than 1/3 of the world's 65 million people with epilepsy (~3.3 million in the U.S.) have seizures that cannot be controlled by medications. Surgery and implanted devices are options for many, but their success depends upon manually mapping epileptic networks, which is only possible for some patients, and poorly standardized. When surgical targets are identified, there is currently no rigorous way to select the best surgical approach. The overall aim of this proposal is to develop rigorous, standardized, quantitative methods to: (1) map epileptic networks from imaging and Stereo EEG (SEEG), (2) pick the best region for resection, ablation or neuromodulation for individual patients from their data and clinical hypotheses, and (3) to determine when focal intervention is unlikely to succeed. These methods would have tremendous positive impact on clinical care. Over the past four years we have made substantial progress towards these goals. We have developed: (1) robust measures derived from subdural intracranial EEG (ECOG) that predict outcome from epilepsy surgery; (2) personalized methods that localize epileptic networks and predict the impact of different interventions on seizure control; (3) tools that predict the path of seizure spread from combined MRI and IEEG. We also have a track record of openly sharing our methods, data, results and code on http: //ieeg.org, to accelerate research. Based upon this work, we now innovate to solve 3 fundamental challenges to translating our work into practice: (1) Guiding SEEG: We must develop new methods that account for the sparser sampling and different philosophy of stereo EEG, which maps a network of connected brain regions and tests clinical hypotheses about where seizures initiate and propagate; (2) Assessing sampling bias and missing information: We will develop methods to determine if electrodes sample all key regions of the epileptic network, to ensure we do not falsely localize due to missing information; (3) Validating in a larger population across centers: In parallel to refining the above methods, we will validate and harmonize our analyses across centers in a large number of patients to harden it for clinical use. In a novel model, we have engaged a group of major surgical epilepsy centers to openly collaborate, standardize methods, aggregate data, and share all algorithms, computer code, data and results on http: //ieeg.org. Our central hypothesis is that our quantitative methods can be standardized across centers, predict outcome from personalized epilepsy surgery, and ultimately be translated to improve clinical care. This work is significant because it merges state of the art network neuroscience, engineering, neurology and neurosurgery to make practical tools to improve and standardize patient care. It also establishes a collaboration between 15 major epilepsy centers to standardize and share data. Finally, this project leverages a thriving collaboration between experts in neurology, computational neuroscience, neurosurgery, neuroimaging and bioengineering at Penn, with a strong track record of clinical translation.