SCH: A Sensing Platform Monitoring Interactions with Daily Objects to Assess Real-World Motor Performance in Stroke Survivors

中风是全球成年人残疾的主要原因。上肢麻痹是最常见的 中风后损伤。脑卒中康复的最终目的是改善患者的运动功能 在其家庭和社区环境中的表现（即，患者实际上是什么）。但目前的 监测患者恢复过程的临床标准仅限于评估患者的运动能力 在临床上观察到（即，患者可以做什么）。腕戴式加速度计已经被 被认为是一种潜在的解决方案，但因提供上肢表现的有限视图而受到批评。 因此，研究和临床社区强调需要一种技术解决方案， 支持更全面地了解中风幸存者的运动表现。 在这项工作中，我们建议开发一种新的多模态传感平台，以监测的重要组成部分， 上肢运动表现：运动的数量、类型和质量。为此，我们引入一个 一种新的传感技术，即身体通道识别（BCID），可以准确， 可靠地跟踪人类与环境的交互，从而跟踪人类行为。在我们的环境中， 日常物品都装有小型、廉价、无电池的BCID标签， 并通过利用人体作为信号与腕戴设备（所谓的读取器）通信 在触觉交互期间传输通道。该系统提供多模态数据，包括对象ID， 交互模式的二元时间序列（接触与非接触），来自可选压力的动力学数据 嵌入在标签中的传感器，以及来自腕戴式传感器上的惯性测量单元的运动学数据。 读者利用从50名中风幸存者和10名健康受试者中获得的数据，我们建议 开发一套独特的机器学习算法来处理这些数据， 与中风康复相关的重要类型的上肢运动，其被进一步处理， 评估所执行动作的质量和数量。最后，我们研究了 在诊所中观察到的运动能力与诊所外的运动表现，这是一个已经被 被认为是中风康复的关键，但由于技术限制而不可行。我们认为， 研究将奠定技术基础，开辟新的研究和临床机会， 关键的科学发现，以改变目前的中风康复实践。

Stroke is the leading cause of disability in adults worldwide. Upper-limb paresis is the most common impairment post-stroke. The ultimate goal of stroke rehabilitation is to improve patients’ motor performance in their home and community settings (i.e., what patients actually do). However, current clinical standards to monitor patients’ recovery process are limited to assessing patients’ motor capacity observed in the clinic (i.e., what patients are capable of doing). Wrist-worn accelerometers have been considered as a potential solution but criticized for providing a limited view of upper-limb performance. Therefor, the research and clinical communities have emphasized the need for a technological solution to support a more comprehensive understanding of stroke survivors’ motor performance. In this work, we propose to develop a novel multi-modal sensing platform to monitor important elements of upper-limb motor performance: the amount, type, and quality of movements. To that end, we introduce a new kind of sensing technology, namely Body Channel Identification (BCID), that can accurately and reliably track human interactions with the environment and, thus, human behaviors. In our setting, everyday objects are instrumented with small, inexpensive, batteryless BCID tags that can be powered by and communicate with wrist-worn devices (so-called readers) by exploiting the human body as the signal transfer channel during tactile interactions. The system provides multi-modal data, including the object ID, binary time-series of interaction patterns (contact vs. no contact), kinetic data from an optional pressure sensor embedded in the tag, and kinematic data from the inertial measurement unit on the wrist-worn readers. Leveraging data obtained from 50 stroke survivors and ten healthy subjects, we propose to develop a unique set of machine learning algorithms to process these data to taxonomically identify important types of upper-limb movements relevant to stroke rehabilitation, which are further processed to assess the quality and amount of movements performed. Finally, we investigate the relationship between the motor capacity observed in the clinic vs. motor performance outside the clinic, a topic that has been deemed critical in stroke rehabilitation but infeasible due to technical limitations. We believe the proposed research will lay the technological groundwork to open up new research and clinical opportunities, leading to key scientific discoveries to transform current practices of stroke rehabilitation.