Alwan Majd, Ledoux Alexandre, Wasson Glenn, Sheth Pradip, Huang Cunjun
Medical Automation Research Center (MARC), P.O. Box 800403, Charlottesville, VA 22908, USA.
Med Eng Phys. 2007 Apr;29(3):380-9. doi: 10.1016/j.medengphy.2006.06.001. Epub 2006 Jul 13.
This paper describes a method that passively assesses basic walker-assisted gait characteristics using only force-moment measurements from the walker's handles. The passively derived gait characteristics of 22 subjects were validated against motion capture gait analysis. The force-moment based heel initial contact detection algorithm have produced a high level of concordance with heel initial contacts detected by a human inspecting the heel marker data sets of the Vicon video capture system. The algorithm has demonstrated 97% sensitivity and 98% specificity with a narrow 95% confidence interval of +/-1% during all experiments, which included five navigational scenarios. Temporal error in detecting the instances of heel initial contacts were within 5.27+/-3.66% of the overall stride time obtained from Vicon when the subjects walked in a straight line, whereas the toe-off instance estimates were within 5.18+/-2.75% of the gait cycle. The errors in determining the duration of stride time, single support, and double support were within 5.86+/-2.49%, 5.24+/-2.29%, and 4.34+/-2.13% of the gait cycle respectively. The stride time estimated, using the method presented here, correlated well with stride time computations based on visual inspection of Vicon's data, Pearson correlation coefficient r=0.86 for straight line segments. However, absolute errors were too high to estimate the single and double support phases with acceptable accuracy. The potential application of the instrumented walker and the method presented here is longitudinal basic gait assessment that can be performed outside of the conventional gait labs.
本文描述了一种仅使用步行器手柄上的力-力矩测量值来被动评估基本步行器辅助步态特征的方法。针对22名受试者通过被动得出的步态特征，与运动捕捉步态分析进行了验证。基于力-力矩的足跟初始接触检测算法与通过人工检查Vicon视频捕捉系统的足跟标记数据集检测到的足跟初始接触具有高度一致性。在包括五种导航场景的所有实验中，该算法的灵敏度为97%，特异性为98%，95%置信区间狭窄，为±1%。当受试者直线行走时，检测足跟初始接触实例的时间误差在Vicon获得的总步幅时间的5.27±3.66%以内，而离地实例估计值在步态周期的5.18±2.75%以内。确定步幅时间、单支撑和双支撑持续时间的误差分别在步态周期的5.86±2.49%、5.24±2.29%和4.34±2.13%以内。使用本文提出的方法估计的步幅时间与基于对Vicon数据的目视检查计算的步幅时间相关性良好，直线段的皮尔逊相关系数r = 0.86。然而，绝对误差过高，无法以可接受的精度估计单支撑和双支撑阶段。本文介绍的仪器化步行器和方法的潜在应用是可以在传统步态实验室之外进行的纵向基本步态评估。
