The goal of this proposal is to develop and validate a practical method to predict epileptic seizures in human temporal and extratemporal epilepsy. Recent data from our laboratory suggest that human partial seizures are associated with a build-up of electrical activity minutes to hours prior to their onset on intracranial EEG (IEEG). The three most promising measures of this pre-ictal build-up are accumulating energy, subclinical seizure-like bursts (chirps), and high frequency epileptiform oscillations. In addition to increasing before seizures, these parameters wax and wane at other times, suggesting that recurrent changes in brain excitability occur repetitively and only proceed to seizures at critical times. By tracking the above three measures in continuous, long-term, multi-channel intracranial EEG (IEEG) data we plan to develop a practical model of how seizures are generated in the epileptic network and will prospectively validate the model's ability to identify periods of increased probability of seizure onset (our definition of """"""""seizure prediction""""""""). Algorithms developed in our laboratory based upon the quantitative features above are currently operating in first generation responsive brain stimulation devices for epilepsy being implanted in about 200 patients in an ongoing clinical trial, with encouraging results. These devices stimulate the brain in response to build-ups of the above quantitative measures in single channels. Optimal performance of these devices will require understanding how these measures develop and spread in the entire epileptic network, and the mechanisms underlying this process. Motivated by the above developments, the specific aims of this proposal are: (1) To meticulously collect, mark and archive a digital database of IEEG studies from a representative population of adults and children with medically resistant temporal and extra-temporal epilepsy, (2) to study the occurrence and duration of the above 3 quantitative measures in all intracranial electrode contacts and prospectively determine their relationship to electrographic seizure onset in continuous, undipped patient data sets; (3) to develop a practical model of seizure generation based upon these findings and prospectively validate its ability to predict seizures. Accomplishing these aims will yield important insight into the mechanisms underlying seizure generation and will be critical to improving the performance of 1st generation reactive epilepsy devices. Our lab will lead an established team of collaborators in this project for data collection, processing and interpretation, with expertise in electrical engineering, neuroscience, clinical epilepsy, neuropathology and statistics at The University of Pennsylvania, The Mayo Clinic and The Georgia Institute of Technology.
Showing the most recent 10 out of 43 publications