Development of TransKinect: A Clinically Robust System for Transfer Assessment

不恰当的移位技术使轮椅使用者容易出现上臂疼痛和损伤。一个 有大量关于转移以避免外力的适当方式的研究，但这些技术 尚未在临床上得到很好的传播。在最近的一项研究中，我们发现高达63%的退伍军人使用 轮椅有许多转移技能缺陷，包括手臂位置不正确，使用手柄和头部- 臀部关系。我们建议开发TransKinect，这是一个用于临床应用的自动传输评估系统 可以帮助治疗师和他们的患者识别不正确的转移动作并提供指导的设置 关于如何提高他们的技术。该系统将基于转让评估表(TAI)。 这是一个有效和可靠的量表，用于评估转移技术的质量，但需要新的治疗师 相当长的时间来学习和判断适当的身体位置和力学。 TransKinect将消除对TAI背景知识和培训的需求，并提供客观的 使用便携式、低成本的无标记运动捕捉传感器测量身体和关节运动。在之前 工作中，我们为TransKinect开发了机器学习分类器，可以区分正确和不正确 使用在传输过程中记录的传感器数据的平均准确率为94%的技术。这些量词 将被嵌入到系统软件中，该系统软件被设计为‘监视’转账，自动计算TAI分数， 实时呈现结果，并向治疗师和他们的 病人。 具体目标是：1)迭代开发TransKinect软件平台，涉及来自 专家小组，2)通过比较系统生成的TAI评分来验证TransKinect是否可用于临床 与专家治疗师的对比3)与一组新手治疗师一起测试TransKinect的可用性；4) 在退伍军人事务部的临床环境中对该系统进行现场测试，以检查其实用性、可用性和有效性。TransKinect将成为 使用在研究的每个目标中收集的数据反复精炼。 对于目标1，将开发TransKinect的原型，并与熟悉 泰。在目标2中，三位专家治疗师和TransKinect将在观看完30名退伍军人后为TAI评分 执行轮椅转运。我们的目标是在两套分数之间达到至少80%的一致性 并将在必要时使用为实现这一目标而收集的数据来完善软件。对于Aim 3，将有10名新手治疗师 使用TransKinect评估模型患者，并提供他们对可用性和感知的反馈 系统的有用性。我们假设治疗师会发现TransKinect易于设置和使用，并且 这增加了他们对转移技术的认识和理解。在目标4中，TransKinect系统 将在Clement J.Zablocki VA医疗中心引入临床实践，我们将在那里跟踪 系统的实用性、可用性和有效性。疗效将通过比较患者的TAI评分来评估 他们在诊所的首次和后续访问中进行了评估。我们希望表明，该系统可以 成功融入标准实践，并对检测转移技术的改进很敏感。

Improper transfer technique predisposes wheelchair users to developing upper arm pain and injuries. A large body of research on the proper ways to transfer to avoid extraneous forces exists but these techniques have not been well disseminated into clinics. In a recent study we found that up to 63% of Veterans who use wheelchairs have many transfer skill deficits including improper arm positioning, use of handgrips, and head- hips relationship. We propose to develop TransKinect, an automated transfer assessment system for clinical settings that can help therapists and their patients to identify improper transfer motions and provide guidance on how to improve their technique. The system will be based on the Transfer Assessment Instrument (TAI) which is a valid and reliable scale used to assess the quality of transfer technique but requires new therapists considerable time to learn and to make judgements on appropriate body positioning and mechanics. TransKinect will eliminate the need for background knowledge and training on the TAI and provide objective measurement of body and joint motions using a portable, low-cost markerless motion capture sensor. In prior work, we developed machine learning classifiers for TransKinect that can differentiate proper from improper techniques with an average accuracy of 94% using the sensor data recorded during a transfer. These classifiers will be embedded in system software designed to `watch' a transfer, automatically compute the TAI score, present the results in real-time and provide education and training recommendations to therapists and their patients. The specific aims are: 1) to iteratively develop the TransKinect software platform involving input from an expert panel, 2) to validate TransKinect for clinical use by comparing the TAI scores generated by the system against those of expert therapists 3) to test the usability of TransKinect with a novice group of therapists and 4) field test the system in a VA clinical setting to examine its utility, usability and effectiveness. TransKinect will be iteratively refined using the data that is collected in each aim of the study. For Aim 1, a prototype of TransKinect will be developed and tested with field experts familiar with the TAI. For Aim 2 three of the expert therapists and TransKinect will score the TAI after watching 30 Veterans perform wheelchair transfers. We aim to achieve at least 80% of agreement between the two sets of scores and will refine the software using data collected in this aim if necessary. For Aim 3, 10 novice therapists wlll use TransKinect to assess a model patient and provide their feedback on the usability and perceived usefulness of the system. We hypothesize that the therapists will find TransKinect easy to setup and use and that it increases their awareness and understanding about transfer technique. In Aim 4 the TransKinect system will be introduced into clinical practice at the Clement J. Zablocki VA Medical Center where we will track the system's utility, usability and effectiveness. Effectiveness will be evaluated by comparing TAI scores on patients who were assessed at an initial and followup visit to the clinic. We hope to show that the system can be successfully integrated into standard practice and is sensitive to detecting improvements in transfer technique.