Development and Acceptability of an Ambient In-Home Activity Assessment Tool for Stroke

1.项目摘要/摘要 在美国和那些没有中风的人中，中风是导致严重、长期残疾的主要原因 锻炼或经常参加活动会使中风复发的风险增加30%。一对一- 由于保险法规和治疗师的原因，One康复课程的数量经常受到限制 (身体和职业)经常开出家庭锻炼计划。这些节目在历史上 遵从率低，患者报告往往存在偏见、不完整或不准确。可穿戴传感器 可以跟踪活动量，但这些传感器的范围有限，无法区分不同的 活动。在家中可以使用深度传感器来检测跌倒并监控老年人的室内步态模式 成年人。其他研究人员使用深度传感器来探测和辨别实验室或模拟实验室中的活动 无任何残疾人口的家庭环境。在这份提案中，每日活动表彰 评估系统(DRAS)将把以前的环境深度传感器工作与新开发的 客观和准确地测量居住在美国的中风患者的活动量和类型的算法 回家。这将通过三个具体目标来完成。将为以下目的开发和改进DARAS算法 使用Preresite深度传感器识别中风患者在厨房环境中的活动。这些 算法将使用来自中风患者(n=10)的实验室测试的真实数据进行训练。我们 将改进基于卷积神经网络(CNN)的算法，以精确分割和 从深度视频的未修剪处理中识别活动。开发的活动识别系统 将在一年内部署在10名中风患者的家中。要确定影响 关于日常生活以及系统和生成的数据的可接受性，将由10个人组成焦点小组 中风。该提案中制定的DRAS将为各种 卒中后干预，并为职业治疗师提供及早发现工作表现下降的能力 并进行干预。

1. Project Summary/Abstract Stroke is the leading cause of serious, long-term disability in the United States and those that do not exercise or engage in regular activity are at a 30% increased risk of experiencing a recurrent stroke. One-on- one rehabilitation sessions are frequently limited in number due to insurance regulations and therapists (physical and occupational) frequently prescribe home-based exercise programs. These programs historically have low adherence rates and patient report can often be biased, incomplete, or inaccurate. Wearable sensors can track amount of activity but these sensors are limited in scope and cannot discern between various activities. Depth sensors can be used in the home to detect falls and monitor in-home gait patterns of well older adults. Other researchers have used depth sensors to detect and discern activities in laboratories or mock home environments with a population without any disabilities. In this proposal, the Daily Activity Recognition and Assessment System (DARAS) will merge prior ambient depth sensor work with newly developed algorithms to objectively and accurately measure the amount and type of activity of people with stroke living at home. This will be completed in three specific aims. The DARAS algorithms will be developed and refined for recognizing activities of people with stroke in the kitchen environment using the Foresite depth sensor. These algorithms will be trained using real-world data from lab-based testing with individuals with stroke (n =10). We will refine the Convolutional Neural Networks (CNN) based algorithm for accurately segmenting and recognizing activities from untrimmed processing of depth videos. The developed activity recognition system will be deployed in the homes of 10 individuals with stroke over the course of 1 year. To determine the impact on daily life and acceptability of the system and generated data, focus groups will be held with 10 individuals with stroke. The DARAS developed in this proposal will provide a novel outcome assessment for a variety of post-stroke interventions and provide occupational therapists the ability to detect declines in performance early on and intervene.