CAREER: Personalizing sensory-driven computerized interfaces to optimize motor rehabilitation

大脑或脊髓受伤的人经常接受体育训练，以恢复日常活动所需的运动能力。为了鼓励人们参与这种“运动疗法”，越来越多地使用了虚拟现实(VR)技术。虚拟现实可以被编程为高度逼真，也可以看起来像一场游戏，以鼓励参与者。尽管如此，使用VR运动疗法的人并不总是比传统方法获得更好的结果，部分原因是它们不是针对每个用户进行个性化的。改进VR治疗需要更好地了解人们对可编程功能的反应。可以提高VR运动疗法的两个特点是训练任务的难度和训练期间提供的指导(反馈)。VR界面可以刺激感官以获得更大的沉浸感，可以使用增强感官反馈(ASF)轻松提供训练指导。ASF的方法包括提供关于一个人在训练期间应该移动的方向和幅度的感觉(例如，视觉、触觉)提示。本项目将研究在适应任务难度和ASF的VR训练过程中，功能障碍如何影响个人表现、生理反应和感知。该项目将通过对高中生进行监督培训来促进教育，以定制开发VR康复应用程序。这些学生将通过在实验室演示中向来自资源不足地区的K-12学生展示他们的工作来进一步参与社区。此外，他们将从护理人员、临床医生以及最重要的是作为最终用户导师的脊髓损伤(SCI)患者那里收到关于他们申请的建设性反馈。该项目将根据表明幸福感和身体准备程度的指标，测试当采用VR训练特征(即任务复杂性、用于指导的ASF)时，运动能力会提高这一假设。该项目的技术目标是：(1)创建回归模型，将VR运动训练期间的表现与感知和生理测量联系起来，并作为不同残疾的函数，以及(2)在使用正强化的自适应控制系统中利用这些模型来个性化VR训练功能。实验将包括SCI患者使用机械臂化身执行VR运动康复任务。该项目关注的是SCI人群，VR运动康复服务不足，但由于SCI可能表现为伴有或不伴有认知障碍的身体残疾，因此是基础科学研究的理想对象。参与者将产生在皮肤表面测量的肌肉激活信号，作为机器学习算法的输入，该算法检测控制化身的命令意图。来自虚拟现实环境、身体传感器和调查的数据将被用来评估训练过程中表现、生理反应和感知的变化。因此，该项目促进以人为中心的培训，在为临床医生提供新的和不同维度的患者特定数据以个性化医疗保健以实现更好功能的过程中，考虑到身体和精神。该项目有可能对用户设备界面如何更好地适应功能产生重要的见解，包括如何提供感官驱动的培训，以便更直观地控制设备(例如，智能手机、车辆、远程控制机器人)。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Persons with an injury to the brain or spinal cord often undergo physical training to recover movement abilities needed for daily activities. To motivate participation in this “motor therapy,” virtual reality (VR) technologies are increasingly used. VR can be programmed to be highly realistic or look like a game to encourage participants. Still, persons who use VR motor therapy do not always achieve better outcomes than traditional methods, partially because they are not personalized to each user. Improving VR therapy requires a greater understanding of how persons respond to programmable features. Two features that can improve VR motor therapy are the difficulty of the training task and the guidance (feedback) provided during training. VR interfaces that stimulate senses for greater immersion can readily provide training guidance using augmented sensory feedback (ASF). The method of ASF involves providing sensory (e.g., visual, haptic) cues about the direction and magnitude in which a person should move during training. This project will examine how impairment in function affects personal performance, physiologic responses, and perceptions during VR training that adapts the task difficulty and ASF. This project will advance education through supervised training of high-school students to custom-develop VR rehabilitation applications. These students will further engage the community by presenting their work at laboratory demonstrations to K-12 students from under-resourced districts. In addition, they will receive constructive feedback about their applications from caregivers, clinicians, and, most importantly, persons with a spinal cord injury (SCI) serving as end-user mentors. The project will test the hypothesis that motor performance improves when adapting VR training features (i.e., task complexity, ASF for guidance) based on measures indicating well-being and physical readiness for training. This project’s technical objectives are: (1) creating regression-type models relating performance to perceptional and physiologic measures during VR motor training and as a function of varying disability, and (2) leveraging these models in an adaptive control system using positive reinforcement to personalize VR training features. Experiments will include persons with SCI performing VR motor rehabilitation tasks with a robot-arm avatar. This project focuses on the SCI population, under-served by VR motor rehabilitation yet ideal for fundamental scientific study since SCI can present as a physical disability with or without cognitive impairment. Participants will generate muscle-activation signals, measured at the skin surface, serving as inputs to a machine-learning algorithm that detects command intention in controlling the avatar. Data from the VR environment, body-mounted sensors, and surveys will be used to assess changes in performance, physiologic responses, and perceptions during training. As such, this project promotes human-centered training that considers both body and mind in providing clinicians with new and diverse dimensions of patient-specific data to personalize healthcare for better function. This project has the potential to generate critical insights into how user-device interfaces are best adapted for function, including how to deliver sensory-driven training for more intuitive control of devices (e.g., smartphones, vehicles, remote-controlled robots).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.