On Line Real Time Seizure Prediction

在线实时癫痫发作预测

• 批准号: 6979177
• 负责人: James Chris Sackellares
• 金额: $ 47.61万
• 依托单位: UNIVERSITY OF FLORIDA
• 依托单位国家: 美国
• 财政年份: 2005
• 资助国家: 美国
• 起止时间: 2005-07-19 至 2010-02-28
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/6979177
• 关键词: brain electrical activity; clinical research; computational neuroscience; computer program /software; electrodes; electroencephalography; epilepsy; human subject; mathematics; patient monitoring device; time resolved data

DESCRIPTION (provided by applicant): We have designed and developed several computer algorithms designed to monitor the spatiotemporal dynamics of the EEC signal and predict seizures. Preliminary data indicates that some algorithms can predict seizures with at least 80% sensitivity (SENS) with a false-prediction rate (FPR) of approximately 0.125 per hour. We propose to evaluate the on-line real-time performance of the algorithm which performed best in our preliminary studies. The algorithm monitors EEG signals from each scalp electrode, iteratively calculates the degree of chaos (values of STLmax) from 10.24 second epochs of signal from each electrode, automatically selects sites likely to participate in the next preictal transition, and predicts a seizure when a preictal transition is detected. It is incorporated into Dynamical Seizure Warning System (DSWS) consisting of a C++ program running on a PC, networked to the clinical recording systems in the Epilepsy Monitoring Unit. The Specific Aims of the project are: (1) determine SENS and FPR of the DSWS when used on-line in real time in long-term EEG recordings with scalp and with intracranial depth electrodes, (2) test the hypothesis that the on-line, real-time performance of the DSWS is better than statistically-based naive prediction algorithms, and (3) evaluate the cause of failed predictions and false predictions generated by the DSWS. This study is designed primarily to evaluate the performance of DSWS in inpatient monitoring applications. However, the results will also be useful for evaluating the potential of this algorithm for other applications, including closed-loop therapeutic devices for the vagal nerve stimulation, or deep brain stimulations. For example, it is likely that such stimulators are more effective if stimulation in activated prior to an impending seizure, and may reduce the amount of stimulation required for seizure control. Information obtained in this study will provide information as to the feasibility of such devices as well as provide information that will be useful in the design of such devices.

描述（由申请人提供）：我们设计并开发了几种计算机算法，旨在监视EEC信号的时空动力学并预测癫痫发作。初步数据表明，某些算法可以预测至少80％灵敏度（SENS）的癫痫发作，错误预测率（FPR）约为每小时0.125。我们建议评估在初步研究中表现最佳的算法的在线实时性能。该算法从每个头皮电极监控EEG信号，迭代地计算从每个电极的10.24秒信号时期的混乱程度（Stlmax值）的程度，自动选择可能参与下一次过渡的站点，并预测在检测到前期过渡时癫痫发作时。它被合并到由PC上运行的C ++程序组成的动力癫痫发警告系统（DSWS）中，该程序与癫痫监测单元中的临床记录系统网络。 The Specific Aims of the project are: (1) determine SENS and FPR of the DSWS when used on-line in real time in long-term EEG recordings with scalp and with intracranial depth electrodes, (2) test the hypothesis that the on-line, real-time performance of the DSWS is better than statistically-based naive prediction algorithms, and (3) evaluate the cause of failed predictions and false predictions generated by the DSWS.这项研究主要旨在评估DSW在住院监测应用中的性能。但是，结果也可用于评估该算法对其他应用的潜力，包括用于迷走神经刺激或深脑刺激的闭环治疗设备。例如，如果在癫痫发作之前被激活的刺激，这种刺激器可能会更有效，并且可能会减少癫痫发作控制所需的刺激量。在本研究中获得的信息将提供有关此类设备可行性的信息，并提供将在此类设备设计中有用的信息。