Abstracts

In the last decade, substantial progress has been made in the study of the human epileptic brain by utilizing concepts and measures from nonlinear dynamics. The hallmark of this research is the ability to predict seizures prior to their clinical or electrographic onset (IEEE TBME 2003; :616-627). As the ability to predict leads to the possibility of control, research in controlling seizures with closed-loop systems is expected to flourish in the near future and will most likely include studies performed in animal models of epilepsy. We have thus applied the concepts from nonlinear dynamics, namely the Largest Short-Term Lyapunov exponent (STL[sub]max[/sub]), to evaluate the effectiveness for real-time seizure prediction in the lithium pilocarpine rat model of chronic epilepsy as a precursor for a closed-loop seizure control system that utilizes deep brain stimulation., Three male Sprague-Dawley rats (300-350g) were stereotaxcially implanted with a customized array of Tungsten depth wire electrodes which included four cortical, two hippocampal, and two thalamic contacts. Chronic epilepsy was established several weeks after an episode of prolonged status epilepticus using the lithium pilocarpine model (3 mmol/kg LiCl followed by 30 mg/kg pilocarpine 24 hrs later). Continuous EEG/video recordings were made during the entire experiment and all data analyzed in real-time. STL[sub]max[/sub] values were calculated for each electrode and entrainment of STL[sub]max[/sub] was evaluated utilizing a pair t-statistic (T-index). True and false predictions were noted with a seizure prediction horizon set at 180 minutes., Continuous EEG data from Rat1 (55 seizures in 166 hours), Rat2 (25 seizures in 384 hours), Rat3 (16 seizures in 739 hours) resulted in a total of 96 seizures in 1289 hours of recording. The seizure prediction results from each rat were: Rat1, prediction rate 35/55=63.4%, 11 false positives (0.0663/hr), mean prediction time=111 minutes; Rat2, prediction rate 22/25=88.0%, 50 false positives (0.1302/hr), mean prediction time=149 minutes; Rat3, prediction rate 10/16=62.5%, 114 false positives (0.1543/hr), mean prediction time=114 minutes. Overall prediction rate- 67/96=69.8%, 175 false positives (0.1358/hr), mean prediction time=125 minutes., While the overall prediction rate of [sim]70% is less than the one reported in human data ([sim]80%), similar false positive rates ([sim]1 every 8 hours of recording), and an extended prediction time were noted. Given these results, prediction of seizures in the lithium pilocarpine rat model of chronic epilepsy appears feasible as a model of chronic epilepsy for testing real-time feedback control systems utilizing thalamic deep brain stimulation., (Supported by Epilepsy Research Foundation of America and Ali Paris Fund for LKS Research, and Barrow Neurological Foundation.)