EAGER: Modeling Network Dynamics in the Epileptic Brain to Develop Translational Tools for Seizure Localization and Detection

点击翻译按钮获取中文摘要

Objective: Epilepsy affects 60 million people worldwide who suffer from recurrent seizures, and 40% of patients do not respond to any drug therapy. These patients would greatly benefit from closed-loop neuro-stimulation therapy to suppress seizures, but the efficacy of such therapy critically depends on whether the stimulus is administered close to the seizure origin (epileptogenic zone, EZ) and immediately prior to or at seizure onset. This program develops novel computational tools for effective EZ localization and seizure onset detection from multi-channel intracranial EEG (iEEG) recordings. Intellectual Merit: The tools are derived by (i) analyzing the dynamics of the brain network as a seizure approaches and (ii) developing a model-based framework that combines multivariate statistics, Bayesian estimation, and optimal control. The tools use iEEG recordings to (1) reconstruct and track the topology of the brain network over time, and (2) identify topological signatures that are specific of the seizure state and uniquely localize the EZ. The rule that detects these signatures from sequential iEEG measurements is adaptive and optimizes the trade-off between specificity and sensitivity by minimizing a cost function of both the detection delay and the probability of false positives. Broader Impacts: Multiple translational impacts will occur at the interface between engineering and neuroscience. First, the proposed tools will allow more accurate EZ localization and resection, more efficient review of iEEG signals, and more effective treatments for seizure suppression (more effective placement of the stimulation electrodes and more efficient neuro-stimulation devices). Overall, these outcomes will reduce the hospitalization time and potentially avoid fatal accidents to epilepsy patients, save lives, extend life-expectancy, and improve the administration of drugs. Also, this program will introduce a transformative detection paradigm that generalizes to any application involving hidden state transition detection relevant to a wide array of disciplines (e.g., early earthquake detection or threats detection). Finally, this program will support the development of courses in multivariate signal processing and statistical modeling at Johns Hopkins University and of outreach activities that will inspire high school students (especially from minorities) from the Baltimore metropolitan area to pursue a career in engineering.