Machine Learning Algorithms to Measure Physical Activity in Children with Cerebral Palsy

用于测量脑瘫儿童身体活动的机器学习算法

• 批准号: 9346616
• 负责人: Margaret Eileen ONeil
• 金额: $ 17.21万
• 依托单位: DREXEL UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2016
• 资助国家: 美国
• 起止时间: 2016-09-07 至 2019-08-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/9346616
• 关键词: Acceleration; Accelerometer; Activities of Daily Living; Adolescent; Algorithms; Behavior; Biological Neural Networks; Brain Injuries; Cerebral Palsy; Child; Child Behavior; Childhood; Classification; Clinical; Comorbidity; Data; Development; Effectiveness; Energy Metabolism; Engineering; Fatigue; Frequencies; Gait; Gait speed; Goals; Health; Health Benefit; Intervention; Knowledge; Laboratories; Learning; Light; Live Birth; Logistic Regressions; Machine Learning; Measurement; Measures; Mechanics; Methodology; Methods; Mission; Modeling; Monitor; Motion; Motor; Movement; Musculoskeletal; Obesity; Operative Surgical Procedures; Orthopedic Surgery procedures; Osteoporosis; Outcome; Outcome Measure; Patients; Pattern; Pattern Recognition; Performance; Physical activity; Physical assessment; Physical therapy exercises; Physically Handicapped; Prevalence; Public Health; Rehabilitation therapy; Research; Research Personnel; Sampling; Science; Signal Transduction; Supervision; Testing; Therapeutic Intervention; Time; Training; United States National Institutes of Health; Work; Youth; base; chronic pain; clinical application; computer science; computerized data processing; cost; experience; fitness; improved; innovation; light intensity; multidisciplinary; neuromuscular; novel; pediatric patients; post intervention; prediction algorithm; primary outcome; rehabilitation science; research study; sedentary; sensor; time use; vigorous intensity; walking speed

PROJECT SUMMARY/ABSTRACT Cerebral palsy (CP) is the most common physical disability of childhood with a prevalence of 2.5 to 3.6 cases per 1000 live births. Inadequate physical activity (PA) and poor fitness are major problems impacting the health and well-being of children with CP. Moreover, low PA may contribute to the development of disabling secondary conditions such as obesity, chronic pain, fatigue, and osteoporosis. Children with CP frequently undergo therapeutic interventions and/or orthopedic surgery to improve their mobility and increase habitual PA. The primary outcome measures used pre-post interventions are typically clinical measures of gross motor function or functional capacity. None of these tests, however, measure PA performance. Accelerometry-based motion sensors are the method of choice for assessing PA in children. Our group has shown that accelerometers provide valid and reliable assessments of ambulatory activity in youth with CP. We have developed and validated CP-specific count thresholds to estimate time spent in sedentary, light-intensity, and moderate-to-vigorous intensity PA. However, there is a knowledge gap on optimal approaches in accelerometer data processing to measure PA in children with CP especially given the misclassification error and because cut-points do not perform well in children with more severe functional limitations. The long-term goal of this project is to improve PA measures in children with CP. The overall objective of this application is to use machine learning in accelerometer data processing to improve PA measures in children with CP. The proposed study will be the first to develop, evaluate, and deploy machine learning algorithms to measure activity type and energy expenditure in children with CP. The specific aims of this project are to: 1) Develop and test machine learning algorithms to predict PA type, walking speed, and energy expenditure in ambulant children and adolescents with CP; 2) compare the accuracy of PA intensity estimates provided by machine learning algorithms to those provided by conventional cut-point methods; and 3) evaluate the performance of the resultant CP prediction models in an independent sample of children with Acquired Brain Injury (ABI). The proposed project is in line with the NIH mission because the resultant prediction models will enable clinicians and rehabilitation professionals to more effectively monitor the PA levels of their patients to improve health and function. Improved objective measures of PA will also enable health researchers to better understand the short-and long-term health benefits of regular PA and impact of PA on adverse health conditions associated with CP.

项目摘要/摘要 脑瘫（CP）是童年最常见的身体残疾，患病率为2.5 每000例活产3.6例。体育锻炼不足（PA）和健身不足是主要的 影响CP儿童健康和福祉的问题。而且，低PA可能 有助于肥胖，慢性疼痛等禁用次要疾病的发展 疲劳和骨质疏松症。 CP的儿童经常接受治疗干预措施和/或 骨科手术以提高其活动能力并增加习惯性PA。主要结果 使用前干预措施的措施通常是总运动功能的临床措施或 功能能力。但是，这些测试均未测量PA性能。 基于加速度计的运动传感器是评估PA的首选方法 孩子们。我们的小组表明，加速度计提供了有效可靠的评估 CP青年的门诊活动。我们已经开发并验证了CP特定的计数 阈值以估算久坐，轻度和中等程度上花费的时间 强度PA。但是，加速度计数据中最佳方法存在知识差距 在CP儿童中测量PA的处理，特别是考虑到错误分类错误和 因为在功能限制更严重的儿童中，切点表现不佳。 该项目的长期目标是改善CP儿童的PA措施。这 该应用程序的总体目的是在加速度计数据处理中使用机器学习 改善CP儿童的PA措施。拟议的研究将是第一个开发的研究， 评估和部署机器学习算法以衡量活动类型和能源 CP儿童的支出。该项目的具体目的是：1）开发和测试 机器学习算法以预测PA类型，步行速度和能源消耗 CP的救护儿童和青少年； 2）比较PA强度估计的准确性 由机器学习算法提供给常规切点方法提供的算法； 3）评估所得CP预测模型的性能 脑损伤儿童（ABI）的样本。 拟议的项目与NIH的任务一致，因为结果预测 模型将使临床医生和康复专业人员能够更有效地监视PA 他们的患者水平以改善健康和功能。改善PA的客观度量将 还使健康研究人员能够更好地了解短期和长期健康益处 常规的PA和PA对与CP相关的不良健康状况的影响。