SBIR Phase I: Particle Filtering Technology for Wearable Medical Sensors

这个小企业创新研究第一阶段项目旨在为可穿戴医疗仪器开发改进的噪声过滤器。近年来，用于监测生理信号的医学传感仪器越来越趋向于可穿戴和非侵入性。然而，由于这些传感器现在是便携式的，它们将暴露于更高水平的噪音和伪影（特别是运动伪影），而不是在受控的临床场景中。因此，这些传感器不能可靠地工作，除非数据经过滤波器的后处理。传统的滤波器（自适应递归滤波器、小波滤波器和其他滤波器）受其通用适用性的限制，不具备必要的性能。为了解决这个问题，流线自动化有限责任公司（SA）和伍斯特理工学院（WPI）将开发基于生理模型的粒子过滤器（PF）。PF已被证明优于所有其他已知的滤波方法，特别是对于非线性系统，如人类生理，具有非平稳和非高斯噪声（运动伪像）。然而，到目前为止，PF尚未用于医学或生物应用。为了实现这一目标，我们提出了一种基于解剖学和生理学概念的状态空间建模方法。在第一阶段，我们将开发和演示基于心血管呼吸系统状态空间模型的粒子滤波方法，以处理可穿戴脉搏血氧仪信号。潜在的应用是巨大的，因为粒子过滤器有潜力为任何生理监测硬件提供鲁棒性和可靠性，但尚未应用于生物信号。该项目的重点是提高可穿戴脉搏血氧仪硬件的稳健性，使其在动态监测中变得有用。该技术应适用于城市和自然灾害地区，在这些地区，必须对多名外伤受害者进行分类和疏散（地震、车祸、爆炸、龙卷风等）。其他应用包括监控远程飞行飞行员、体育锻炼监测、手术和麻醉、睡眠呼吸暂停、慢性心血管或呼吸系统疾病患者，以及在恶劣环境下的远程监控，如高空救援队、消防员和深海潜水。由于颗粒过滤技术不仅限于脉搏血氧测定，还存在许多其他应用，为系统和测量提供合适的数学模型。这些例子是检测胎儿心电图微弱，肾透析监测，无创血糖监测，和肌电图过滤。

This Small Business Innovation Research Phase I project is aimed at developing improved noise filters for wearable medical instrumentation. Recently, medical sensing instrumentation for the monitoring of physiological signals has become increasingly wearable and noninvasive. However, because these sensors are now portable they will be exposed to higher levels of noise and artifacts (especially motion artifacts) than in controlled clinical scenarios. As a result, these sensors cannot perform reliably unless data is post-processed by a filter. Conventional filters (adaptive-recursive, wavelet, and others) are limited by their generic applicability and do not have the necessary performance. To address this, Streamline Automation, LLC (SA) and Worcester Polytechnic Institute (WPI) will develop particle filters (PF) based on physiological models. PF have been shown to outperform all other known filtering methods, especially for nonlinear systems, such as human physiology, with non-stationary and non-Gaussian noise (motion artifacts). However, so far PF have not been used in medical or biological applications. To make this possible, we propose a state-space modeling approach based on anatomical and physiological concepts. In Phase I, we will develop and demonstrate the particle filtering approach based on a cardiovascular-respiratory system state-space model to process wearable pulse oximeter signals.Potential applications are vast because particle filters have the potential to deliver robustness and reliability to any physiological monitoring hardware but have not yet been applied to biosignals. The focus of this project is increasing the robustness of the wearable pulse oximeter hardware such that it becomes useful in ambulatory monitoring. This technology should be applicable in urban and natural disaster areas where multiple traumatic injury victims must be triaged and evacuated (earthquakes, car accidents, explosions, tornadoes, etc). Other applications include monitoring of long-distance flight pilots, physical exercise monitoring, surgery and anesthesia, sleep apnea, patients with chronic cardiovascular or respiratory conditions, and remote monitoring under austere environments such as high altitude rescue teams, firefighters, and deep sea diving. Since particle filtering technology is not limited to pulse oximetry, a host of other applications exist, provided suitable mathematical models for the system and measurement are developed. Examples of these are the detection of faint fetal electrocardiogram, kidney dialysis monitoring, non-invasive glucose monitoring, and electromyogram filtering.