Online Prediction of Gait Related Trips Post-Stroke

摘要 对于从中风中恢复的人来说，受伤的跌倒通常是由于绊倒或“内在原因”造成的。 走路时绊倒(即摆动的脚接触地面)。发展有效下落的主要障碍 预防策略是无法事先可靠地确定是否会发生与步行相关的旅行。秋千 然而，肢体运动是由后期的站姿运动学决定的。因此，我们建议开发一种新的推理 基于站姿阶段运动学的系统，可以准确和可靠地实时预测跳闸 马上就要发生了。因此，如果预测脚将触地，该算法将通知需要哪些步骤 对摆动肢体的轨迹进行干预。目前预防跌倒的方法教会了人们对跌倒的反应 中风后的绊倒或训练，以最大限度地减少与跌倒相关的损害(例如，力量、平衡、 只读存储器)。尽管预防性培训可以减少原本健康的老年人的内在旅行， 自主性肌肉激活的缺陷限制了中风后个人进行这种训练的有效性。而不是 使用传统的反应式方法，我们打算开发开发一个 主动、集成、前馈控制器，用于通知未来的工程方法(例如，多通道 电刺激器、外骨骼/外骨骼)以适当地干预摆动肢体的轨迹，仅当 这是必要的。这里提出的工作是确定我们可以成功预测TRIPS的必要的第一步 准确并有足够的时间进行适当的干预。为了实现这一目标，我们将追求两个具体的目标 目标。在具体目标1中，我们将使用非环境干扰物来增加参与者的可能性 在跑步机和地面上行走时，体验一次内在产生的旅行。到时候我们会的 使用所记录的肢体运动学来选择用于开发新型推理预测系统的特征集。 该算法将进行离线评估，以确定其在将步骤分类为Trips或 非旅行。在目标2中，我们将在TRIP和非TRIP的实时分析中对所开发的推理系统进行评估 走路时的步数。同样，在线系统将在步行试验期间进行准确性和速度评估。在… 这个项目的结论是，我们将有一个健壮的方法来检测即将到来的旅行，以“选择性” 对摆动肢体轨迹的干预。这项工作有望对步行领域产生巨大影响 中风后的康复和其他与旅行相关的跌倒风险人群。特别是，成功完成了 该项目将建立从被动预防跌倒到主动绊倒和预防跌倒的范式转变。

Abstract For individuals recovering from a stroke, injurious falls often occur due to a stumble, or “intrinsically generated” trip (i.e., the swinging foot contacts the ground), while walking. A major barrier toward developing effective fall prevention strategies is an inability to determine reliably, in advance, that a walking-related trip will occur. Swing limb motion, however, is dictated by late stance kinematics. Therefore, we propose to develop a novel inference system, based on stance phase kinematics, that can accurately and reliably predict, in real-time, that a trip is about to occur. Thus, if the foot is predicted to strike the ground, the algorithm will inform which steps require intervention in the swing limb's trajectory. Current approaches to fall prevention teach reactive responses to a trip or train individuals post-stroke to minimize the impairments associated with falls (e.g., strength, balance, ROM). Although preventive training can reduce intrinsically generated trips for otherwise healthy older adults, deficits in voluntary muscle activation limits the efficacy of such training in individuals post-stroke. Rather than using a conventional reactive approach, we intend to develop the preliminary tools needed to develop a proactive, integrated, feed-forward controller to inform future engineering approaches (e.g., a multi-channel electrical stimulator, exoskeleton/exosuit) to appropriately intervene in the swing limb's trajectory, only when necessary. The work proposed here is a necessary first step to determine that we can successfully predict trips accurately and with sufficient time to intervene appropriately. To accomplish this goal, we will pursue two Specific Aims. In Specific Aim 1, we will use non-environmental distractors to increase the likelihood of participants experiencing an intrinsically generated trip while walking on both a treadmill, as well as overground. We will then use the recorded limb kinematics to select a feature set for development of a novel inference prediction system. This algorithm will be evaluated offline to determine its accuracy and speed in classifying steps as either trips or non-trips. In Aim 2, we will evaluate the developed inference system in a real-time analysis of trip and non-trip steps during walking. Again, the online system will be evaluated for accuracy and speed during walking trials. At the conclusion of this project, we will have a robust method of detecting an upcoming trip for “selective” intervention in swing limb trajectory. This work promises to have a tremendous impact on the field of walking recovery post-stroke and for other populations at risk for trip related falls. In particular, successful completion of this project will establish a paradigm shift from reactive fall prevention to proactive trip and fall prevention.