Calibrating free-living physical activity characteristics across functionally-limited populations using machine-learned accelerometer approaches

项目概要/摘要 五分之一的美国成年人被认为患有残疾/损伤，这是一种复杂且多方面的疾病 影响运动模式，相关医疗费用每年超过 3000 亿美元。精确和 准确评估残障人士的体力活动 (PA) 和久坐行为 (SB)/ 损伤对于准确测量 PA/SB 患病率和基于行为的有效性至关重要 PA/SB 干预措施，并充分阐明 PA/SB 剂量-反应健康关系。科学进步已经 在这一领域，先进的分析和数据处理技术应用于可穿戴设备 来自实验室校准研究的加速度计。科学上需要扩展校准研究 固定持续时间的实验室模拟日常生活活动到通过自然观察进行自由生活校准 以及加速度计算法训练和验证。本提案的目标填补了这一重要的科学问题 知识差距。具体目标是：1）评估和完善机器学习算法来预测能量 24 小时呼吸热量计停留期间的费用和活动类型； 2) 验证机器学习加速度计 具有现场导出、视频记录直接观察的算法； 3) 验证机器学习算法 使用双标记水技术。我们高素质的研究团队将通过以下方式实现上述目标： 使用简短的可翻译功能测试将运动障碍人群分为健康组、 上半身损伤、下半身损伤以及上半身和下半身损伤。最佳实践免费- 然后，将使用实时校准协议来训练、完善和评估特定于集群的功能 用于预测活动能源成本、活动类型、活动转换和活动的加速计算法 域。这些拟议研究的结果将首次提供创新的、可转化的 对残疾人和运动障碍人士的自由生活 PA/SB 进行分类和评估的方法。

PROJECT SUMMARY/ ABSTRACT One in 5 U.S. adults are thought to be living with a disability/impairment, a complex and multifaceted condition affecting movement patterns with related medical care costs exceeding $300 billion annually. Precise and accurate assessment of physical activity (PA) and sedentary behavior (SB) in individuals with disabiliy/ impairment is essential to accurately measure PA/SB prevalence rates and effectiveness of behavioral based PA/SB interventions, and to fully elucidate PA/SB dose-response health relationships. Scientific progress has been made in this area with advanced analytics and data processing techniques applied to wearable accelerometers from laboratory calibration studies. There is a scientific need to extend calibration studies from fixed-duration laboratory simulated activities of daily living to free-living calibrations with natural observation and accelerometer algorithm training and validation. The aims of this proposal fill this essential scientific knowledge gap. The specific aims are: 1) To evaluate and refine machine-learned algorithms to predict energy cost and activity type during a 24-hr respiratory calorimeter stay; 2) To validate machine-learned accelerometer algorithms with field-derived, video-recorded direct observation; and 3) To validate machine-learned algorithms using the doubly labeled water technique. Our highly qualified research team will address the above aims by using brief translatable functional tests to cluster movement-impaired populations into groups of healthy, upper-body impairment, lower-body impairment, and upper- and lower-body impairment. Best practice free- living calibration protocols will then be used to train, refine, and evaluate functional clustered-specific accelerometer algorithms for predicting activity energy cost, activity type, activity transitions, and activity domain. The results of these proposed studies will for the first time provide an innovative and translatable approach to categorize and assess free-living PA/SB in persons with disability and movement impairment.