Developing a mobile seizure alert device using non-invasive physiological measure

描述（由申请人提供）：目前，护理人员干预是缓解与糖尿病相关不良事件的唯一方法，并且很少有日常间接监测的选择。用于非临床环境的方法包括夜间床传感器和基于加速度计的传感器，但是这些传感器具有低灵敏度（例如，仅检测夜间或强直阵挛发作）和高误报率。RTI国际和儿童国家医学中心（CNMC）的研究人员提出共同努力开发一种新的癫痫发作警报系统，用于日常监测和护理人员警报。该系统将针对癫痫发作期间自主神经系统（ANS）活动升高引起的有据可查的生理效应，这些效应可以用不显眼的舒适传感器测量。将使用多传感器方法来提高检测所有全身性癫痫发作和某些类型的部分性癫痫发作（不包括失神和简单部分性癫痫发作）的灵敏度和精确度。总体性能目标是证明在95%的时间内可以识别出严重癫痫发作，且假事件率低于10%。从16名受试者中收集的初步数据表明，使用多传感器方法成功检测癫痫发作的可能性很大。通过减少响应时间和消除持续观察的需要，该系统可以通过减少与癫痫相关的伤害和死亡数量，提高生活质量，增加患者和护理人员的独立性，并降低治疗成本，对癫痫社区产生重大和可衡量的影响。研究计划讨论了三个具体目标。目的1：开发和验证癫痫发作检测算法与临床数据。假设可以使用ANS升级的多个生理指标（包括心率和变异性、呼吸、骨骼肌活动和出汗的变化）以高灵敏度和精确度检测显著癫痫发作。将开发和验证一种自动检测算法，该算法使用多传感器数据融合和模式识别方法对癫痫发作和非癫痫发作状态进行分类。目标2：开发综合原型癫痫警报装置。将开发一个完全集成的原型系统，包括一个紧凑的可穿戴监测设备和一个远程护理人员警报单元，该单元基于使用定量临床结果、潜在最终用户反馈和最先进技术进步的稳健方法。目标3：在临床和住宅环境中的原型。在CNMC的临床研究期间，将对癫痫发作警报系统原型的性能和总体可用性进行测试。45对患者-护理人员将在临床测试中使用该系统来评估检测性能，然后进行家庭测试，以评估设备在现实环境中日常使用的有效性、实用性和舒适性。

DESCRIPTION (provided by applicant): Currently, caregiver intervention is the sole method of mitigating seizure-related adverse events, and there are few options for indirect monitoring on a daily basis. Approaches for non-clinical settings have included nocturnal bed sensors and accelerometry-based sensors, but these have suffered from low sensitivity (e.g., only detects nocturnal or tonic-clonic seizures) and high false-alarm rates. Investigators at RTI International and Children's National Medical Center (CNMC) propose a joint effort to develop a novel seizure alert system for daily monitoring and caregiver alert. This system will target the well-documented physiological effects due to elevated activity of the autonomic nervous system (ANS) during seizures that can be measured with unobtrusive, comfortable sensors. A multi-sensor approach will be used to increase sensitivity and precision for the detection of all generalized seizures and some types of partial seizures, excluding absence and simple partial. The overall performance objective is to demonstrate that significant seizures can be identified 95% of the time with a false event rate of less than 10%. Preliminary data collected from 16 subjects suggest that successful detection of seizures with a multi-sensor approach is highly probable. By decreasing response time and eliminating the need for constant observation, this system could have a substantial and measurable impact on the epilepsy community by decreasing the number of seizure-related injuries and deaths, improving quality of life, increasing independence for both patients and caregivers, and reducing the cost of treatment. The research plan discusses three specific aims. Aim 1: Develop and validate seizure detection algorithm with clinical data. It is hypothesized that significant seizures can be detected with hig sensitivity and precision using multiple physiological indicators of ANS escalation, including changes in heart rate and variability, respiration, skeletal muscle activity, and sweating. An automated detection algorithm will be developed and validated that uses a multi-sensor data fusion and pattern recognition approach to classify seizure and non-seizure states. Aim 2: Develop integrated prototype seizure alert device. A fully integrated prototype system will be developed, including a compact, wearable monitoring device and a remote caregiver alert unit based on robust methodology using quantitative clinical results, feedback from potential end users, and state-of-the-art technological advancements. Aim 3: Validate prototype in clinical and residential settings. The prototype seizure alert system will be tested for performance and overall usability during a clinical study at CNMC. Forty-five patient-caregiver pairs will use the system in clinical testing to assess detection performance, followed by at-home testing to assess device effectiveness, practicality, and comfort for daily use in a realistic environment.