地面反力與人體慣性參數之誤差對步態分析結果的影響

Abstract

目前步態分析己被廣泛運用於人體神經骨骼肌肉系統疾病之診斷以及治療的規劃與評估。而步態分析主要利用運動學、人體測計學與測力板資料間接求得人體下肢各關節所承受之作用力。因此，測力板所量測力量與壓力中心是否精準，人體測計學所提供各肢段之質量、質心與慣性矩是否準確，對於步態分析之研究結果有極大的影響。本研究研製一台對固定式測力板作靜態、動態校正的校正器，校正器重量輕、體積小且裝有移動輔助輪，所以可快速移至實驗室作現地校正。校正器固定方式採用吸盤吸附在實驗室地板上因此架設容易，對實驗室建築物無侵入式破壞。校正器在施力點定位與施力大小的控制是採用PC-based 控制器，所以準確性高且快速。本研究利用靜態校正測試作為類神經網路訓諫資料，並將測力板量測力量與壓力中心作修正補償，在垂直力方向其力量誤差平均值百分比在校正前是0.38%，校正後降為0.00%；在壓力中心X、Y軸方向其位置誤差平均值在校正前1.37mm、1.15mm，校正後降為0.02mm、0.04mm。在動態校正方面，在垂直力方向其力量誤差平均值百分比在校正前是-0.19 %，校正後降為-0.03 %；在壓力中心X、Y軸方向其位置誤差平均值在校正前-0.50mm 、0.95mm，校正後降為-0.01mm、 -0.11 mm 。本研究利用測力板校正器對自行研製可測力量跑步機施以垂直力負載校正，並運用類神經網路校正方法來修正跑步機所量測的力量與壓力中心之誤差。在垂直力方向其力量誤差平均值百分比在校正前是0.82%，校正後降為0.01%；在壓力中心X、Y軸方向其位置誤差平均值在校正前1.59mm、0.71mm，校正後降為0.07mm、-0.06mm 。 目前人體測計學不論是利用屍體或侵入性的方法在道德上均不適合兒童，而少數非侵入性方法則因操作不易、設備取得困難、成本過高等因素無法適用於例行臨床步態分析實務與研究。所以本研究利用動作追蹤系統量測各肢段之空間位置及測力板量測力量與壓力中心，再運用最佳化方法來建立個人化人體測計學資料，其中包含各肢段之幾何模擬、各肢段之質量、質心與慣性矩。本研究受試者選取12位健康成人(24□2 yrs; 69□8 kg; 178□5 cm)及20位健康兒童(9□3 yrs; 31□10 kg; 130□9 cm)。在靜態量測時，雙腳站立於測力板上，且擺20種不同姿勢；在動態測試時，受試者則採屈膝下彎動作。本研究方法將求得人體測計學資料與Dempster(1955)、Cheng (2000)人體測計學資料文獻值代入本研究中之人體數學模型作比較。在靜態準確性之壓力中心評估方面，成人壓力中心誤差平均值本研究方法小於5mm內，而文獻方法在11mm~19mm之間；兒童壓力中心誤差平均值本研究方法小於4mm內，而文獻方法在15mm~25mm之間。在動態準確性之壓力中心評估方面，成人壓力中心誤差平均值本研究方法為9.4mm內，而文獻方法在20.6mm~27.9mm之間；兒童壓力中心誤差平均值本研究方法為7.9mm內，而文獻方法在24.8mm~31.1mm之間。在動態準確性之垂直方向地面反作用力評估方面，本研究方法與文獻方法在成人與兒童垂直方向地面反作用力之誤差平均值是相近的。本研究成功發展一套非侵入性、快速、低成本、準確且適合各種體型、性別及年齡的活體個人化量測學資料測量方法，並用以建立我國成人與6~12歲兒童人體測計學資料庫，包括各肢段質量、質量中心及轉動慣量等資料，以供臨床步態及動作分析之需。

Clinical gait analysis is the process of using quantitative information, including kinematic, kinetic and anthropometric data to aid in understanding the etiology of gait abnormalities. It has been widely used in the diagnosis of patients with neuromusculoskeletal pathology, subsequent planning and evaluation of treatment. In human motion analysis, the kinetic data are usually obtained from forceplates mounted on the ground. Therefore, in situ calibration of the forceplate is necessary to improve the accuracy of the measured ground reaction force (GRF) and center of pressure (COP). The current study developed a small device (160 x 88 x 43 cm) with a mass of 50 kg, equipped with auxiliary wheels and fixing suction pads for rapid deployment and easy set-up. A PC-based controller enabled quick movement and accurate positioning of the applied force to the calibration point. After correction by an artificial neural network (ANN) trained with the static data from 121 points, the mean errors for the vertical GRF were all reduced from a maximum of 0.38 % to less than 0.00 %. Those for the X and Y components of COP were all reduced from a maximum of about 1.37 and 1.15 mm to less than 0.02 and 0.04 mm, respectively. For dynamic calibration, the mean errors for the vertical GRF were reduced from a maximum of -0.19 % to less than -0.03 %, while those for the X and Y components of COP were reduced from a maximum of -0.50 and 0.95 mm to less than -0.01 and -0.11 mm. The results suggested that the calibration device with the ANN method will be useful for obtaining more accurate GRF and COP measurements. Thereafter, the device was used to calibrate our newly developed instrumented treadmill to measure GRF on the treadmill during successive cycles of gait. By the same error analysis and neural network methods, the measured GRF and center of pressure (COP) can be calibrated to reduce the errors. The results of calibration indicated that mean errors for the vertical GRF from a maximum of 0.82 % to less than 0.01 %, while those for the X and Y components of COP were reduced from a maximum of 1.59 and 0.71 mm to less than 0.07 and -0.06 mm. Correct anthropometric data is also needed for accurate calculation of the motion data. Currently, anthropometric data are mostly obtained from studies on adult cadavers because no data exist for the children between 6 to 12 years of age. However, methods using cadavers or invasive techniques are not suitable for children. Noninvasive methods are either too difficult or too expensive to be used routinely in clinical settings. The current study therefore aimed to develop a noninvasive, fast, cost-effective and accurate method for the estimation of the anthropometric data of subjects with different ages. We proposed an optimization-based, non-invasive, radiation-free method for estimating subject-specific body segment inertial properties (BSIPs) by using a motion capture system and two forceplates. Twelve healthy adult subjects (24□2 y/o; 69□8 kg; 178□5 cm) and twenty children (9□3 yrs; 31□10 kg; 130□9 cm) were recruited in this study. Firstly, a three-dimensional custom-made model of the human body was developed for the simulation of the segment geometry; the estimation of the mass, center of mass and second moment of inertia of the segments and the whole body. Then the subject was asked to stand in twenty different postures for static test, and to perform squatting for dynamic test. The static and dynamic tests were used to customize the model to the subject with optimization method, and the subject-specific anthropometric data were the calculated consequently. The performance of the current method was compared to two commonly used predictive methods (Dempter, 1955 and Cheng ,2000) in terms of the errors of the calculated COP and ground reaction force (GRF) using the corresponding predicted BSIPs. During stationary standing postures, the mean COP errors were less than 4 and 5 mm for the child and adult groups respectively, while those from the existing comparative methods ranged from 11 to 19 mm and 15 to 25 mm for these two groups respectively. During dynamic activities, mean COP errors from the current method were less than 7.9 and 9.4 mm for the child and adult groups respectively, while those from the existing methods ranged from 24.8 to 31.1 mm and 20.6 to 27.9 mm for these two groups respectively. In evaluation of the accuracy in vertical GRF during dynamic test, the mean error of vertical GRF from the current method showed similar values to the existing methods. The results showed that the current method was capable of producing estimates of subject-specific BSIPs that predicted accurately the important variables in human motion analysis during static and dynamic activities. In conclusion, this optimization-based and accurate method was developed for the estimation of the anthropometric data of subjects with different age groups for clinical gait or motion analysis. Being non-invasive and using standard motion laboratory equipment, the current method would be useful for building up the anthropometric data of adults and children in Taiwan.