Towards Neural Control of Artificial Legs: Design of a Real-Time Fusion-based Neu

描述（由申请人提供）：计算机化和动力假肢的进步显示出极大的希望，使下肢截肢者能够在平地行走之外进行多功能活动。然而，由于缺乏神经控制，这些假肢不足以让用户在活动之间进行无缝转换。为了“告诉”假体预期的运动，用户必须进行额外的身体运动或使用远程密钥卡，这既麻烦又不坚固。通过神经控制接口直接从用户那里获得决策对于提供计算机化人工腿的准确、直观控制至关重要。我们的长期目标是开发一种神经控制的人工膝关节和/或踝关节，以改善计算机化人工腿的功能和下肢截肢者的生活质量。为了实现这一目标，我们建议开发一个强大的神经-机器接口，可以实时识别用户的预期下肢任务。一个功能性的嵌入式神经接口系统将在本项目结束时交付，这可能会在计算机化人工腿的设计中开始一个完整的范式转变。这项资助的具体目标是：目标1：开发一种神经接口算法，可以准确和响应地解码用户预期的下肢任务和任务转换。目标2：在实时嵌入式硬件上实现目标1中设计的算法。目的3：评价膝关节离断或经股截肢受试者的实时神经接口系统。我们提出了一个神经机械融合为基础的接口设计的算法（目的1）的发展。该算法将整合通过肌电图（EMG）记录收集的神经肌肉控制信息与来自假体的机械反馈，以提高识别用户意图的准确性。提出了一种相位相关的模式识别策略，以保证系统的快速响应，以满足实时应用的要求。传感器故障检测器和有限状态机等附加组件将被设计用于增强系统的鲁棒性。所设计的算法将在由自约束仪表腿和嵌入式系统组成的实时测试硬件上实现（Aim 2）。数据结构和程序将被优化，以充分利用嵌入式体系结构和多级存储器层次结构的实时操作。最终的实时神经-机器接口将在膝关节离断或经股截肢的患者中进行评价，这些患者的水平较高且具有挑战性（目标3）。 公共卫生相关性：为人工腿的神经控制开发的神经-机器接口将导致改善受损肢体的功能使用，减少残疾，并提高下肢截肢患者的生活质量。

DESCRIPTION (provided by applicant): Advances in computerized and powered artificial legs show great promise to permit persons with lower limb amputations to perform versatile activities beyond level ground walking. These prostheses are, however, inadequate for users to perform seamless transitions between activities due to the lack of neural control. To "tell" the prosthesis the intended movement, the user must make extra body motions or use a remote key fob, which are both cumbersome and not robust. Obtaining decisions directly from the user through a neural control interface is crucial to providing accurate, intuitive control of computerized artificial legs. Our long-term goal is to develop a neural-controlled artificial knee and/or ankle to improve the function of computerized artificial legs and the quality of life of people with lower limb amputations. Towards this goal, we propose to develop a robust neural-machine interface that can recognize the user's intended lower limb tasks in real-time. A functional, embedded neural interfacing system will be delivered at the end of this project that may start a complete paradigm shift in the design of computerized artificial legs. The specific aims of this grant are: Aim 1: Develop a neural interface algorithm that accurately and responsively decodes the user's intended lower limb tasks and task transitions. Aim 2: Implement the algorithm designed in Aim 1 on real-time embedded hardware. Aim 3: Evaluate the real-time neural interfacing system on subjects with knee disarticulation or transfemoral amputations. We propose a neural-mechanical-fusion-based interfacing design for the development of the algorithm (Aim 1). The algorithm will integrate the neuromuscular control information gathered through electromyographic (EMG) recordings with mechanical feedback from the prosthesis to achieve improved accuracy for identifying user intent. A phase-dependent pattern recognition strategy is proposed to ensure a fast system time response for real-time application. Additional components such as sensor fault detectors and a finite-state machine will be designed to enhance the system robustness. The designed algorithm will be implemented on real-time testing hardware composed of a self-constrained instrumented leg and an embedded system (Aim 2). The data structures and programs will be optimized to make the best use of the embedded architecture and the multilevel memory hierarchy for real-time operation. The finalized real-time neural- machine interface will be evaluated on patients with knee disarticulation or transfemoral amputations, which are high and challenging levels (Aim 3). PUBLIC HEALTH RELEVANCE: The neural-machine interface developed for neural control of artificial legs will lead to improved functional usage of impaired limbs, reduced disability, and improved quality of life of patients with lower limb amputations.