Automated Seizure Detection Following Nerve Agent Exposure

描述（由申请人提供）：神经毒剂（NA）攻击后的癫痫发作会引起时间紧迫的紧急情况，可能导致猝死或严重发病。生存取决于早期和积极的治疗。癫痫发作在体检中可能不明显，急救人员通常无法解释脑电图。因此，如RFA中所述，迫切需要自动EEG解释和癫痫样活动检测。为了满足这一需求，我们组建了一个跨学科团队，来自无限生物医学技术有限责任公司，美国陆军化学防御医学研究所（USAMRICD）和约翰霍普金斯医学院。该团队在NA诱导的神经病理学、癫痫发作的临床管理、定量神经诊断和硬件设计方面拥有丰富的经验。基于这一经验，我们建议开发一种便携式系统，用于自动检测癫痫发作。它包括一个无线头带和一个手持设备，该设备具有化学发作矢量（CSV）算法。CSV是一种基于5种光谱和时间分析的“数字指纹”，专门用于识别癫痫发作形态。该组合方法创建5维向量以可靠地分类和区分癫痫发作波形。噪声降低和伪影抑制被纳入在个人算法的水平，并在5维空间中的个人分析的同时考虑。在第一阶段，我们将评估CSV检测癫痫发作的能力。鉴于研究患者的神经毒剂暴露是不可行的，我们将采用两种互补的模型。首先，我们将使用一个已建立的灵长类动物模型，其中包括暴露于神经毒剂。其次，我们将使用一个涉及癫痫发作患者的非NA相关病因的人类模型。我们将使用受试者操作特征（ROC）曲线研究单个组件算法和组合CSV的性能。阶段I的主要里程碑是基于ROC分析创建最佳化学发作检测算法。这一快速通道应用的第二阶段涉及进一步完善CSV，以产生一个三级指标。我们将从发作周活动（黄色）和正常EEG（绿色）中区分癫痫发作（红色）。CSV将在灵长类动物模型和人类癫痫患者的前瞻性评价中进行测试。该技术将被封装到一个系统中，该系统具有自粘电极头带和手持设备。最后，整个系统将在现场化学应急演习中在现实环境中得到验证。神经毒剂暴露是一种严重的威胁，可导致脑损伤和死亡。我们将开发一个自动化系统来诊断大脑中的EEG癫痫活动，以帮助照顾攻击受害者。

DESCRIPTION (provided by applicant): Seizures following a nerve agent (NA) attack induce a time-critical emergency which can lead to sudden death or severe morbidity. Survival is dependent upon early and aggressive therapy. Seizures may not be evident from physical examination, and emergency personnel typically cannot interpret the electroencephalogram. Thus, as noted in the RFA, there is a pressing need for automated EEG interpretation and detection of epileptiform activity. In order to respond to this need, we assembled an interdisciplinary team from Infinite Biomedical Technologies, LLC, the United States Army Medical Research Institute of Chemical Defense (USAMRICD), and the Johns Hopkins School of Medicine. Together, the team has broad experience with NA-induced neuropathology, clinical management of seizures, quantitative neurodiagnostics, and hardware design. Based on this experience, we propose the development of a portable system for automated seizure detection. It includes a wireless headband and a handheld device which features a Chemical Seizure Vector (CSV) algorithm. The CSV is a "digital fingerprint" based on 5 spectral and temporal analyses specifically selected to recognize seizure morphology. This combinatorial method creates a 5-dimensional vector to reliably classify and discriminate seizure waveforms. Noise-reduction and artifact rejection are incorporated both at the level of the individual algorithm and upon simultaneous consideration of the individual analyses in 5-dimensional space. During Phase I we will evaluate the ability of CSV to detect the presence of seizures. Given that it is not feasible to study nerve agent exposure in patients, we will employ two complementary models. First, we will use an established primate model which incorporates exposure to a nerve agent. Second, we will use a human model involving seizure patients of a non-NA-associated etiology. We will investigate the performance of the individual component algorithms and the combined CSV using Receiver Operator Characteristic (ROC) curves. The primary milestone for Phase I is creation of an optimal chemical seizure detection algorithm based on the analysis of the ROCs. Phase II of this Fast-track application involves further refinement of CSV to produce a three-level indicator. We will discriminate seizures (RED) from peri-ictal activity (YELLOW) and the normal EEG (GREEN). CSV will be tested in a prospective evaluation in the primate model and in human seizure patients. The technology will be packaged into a system which features a self-adhesive electrode headband and a handheld device. Finally, the entire system will be validated in a realistic setting during live chemical emergency drills. Nerve agent exposures are a serious threat and can cause brain injury and death. We will develop an automated system to diagnose EEG seizure activity in the brain to help care for attack victims.