Association of Machine Learning-Based Predictions of Medial Knee Contact Force With Cartilage Loss Over 2.5 Years in Knee Osteoarthritis
Association of Machine Learning-Based Predictions of Medial Knee Contact Force With Cartilage Loss Over 2.5 Years in Knee Osteoarthritis
复制标题
DOI：
10.1002/art.41735
复制
发表时间：
2021-08-06
影响因子：
13.3
通讯作者：
Maly, Monica R.
中科院分区：
文献类型：
--
作者：
Brisson, Nicholas M.;Gatti, Anthony A.;Maly, Monica R.
Objective.尚未研究体内膝关节负荷预测与纵向软骨变化之间的关系。我们进行了这项研究，以开发一个方程来预测内侧胫股接触力（MCF）峰值在步行过程中的人与膝关节内固定植入物，并应用此方程来确定预测的MCF峰值和软骨损失之间的关系，膝关节骨关节炎（OA）。在患有膝关节OA的成人中（39名女性，8名男性;平均年龄61.1 ± 6.8岁），进行基线生物力学步态分析，并使用磁共振成像确定2.5年内胫骨内侧软骨体积的年变化（mm（3）/年）。在使用测力胫骨假体的患者的单独样本中（3名女性，6名男性;平均+/- SD年龄70.3 +/- 5.2岁），使用步态数据加上体内膝关节载荷开发一个方程，使用机器学习预测MCF峰值。然后将该方程应用于膝关节OA组，并确定预测的MCF峰值与软骨体积年变化之间的关系。用步态速度、膝关节内收力矩峰值和膝关节垂直反力峰值预测MCF峰值效果最好（均方根误差132.88N; R-2 = 0.81，P < 0.001）。在膝关节OA受试者中，预测的MCF峰值与软骨体积变化相关（R-2 = 0.35，beta =-0.119，P < 0.001）。使用机器学习开发了一种新的方程，用于从外部生物力学参数预测MCF峰值。在膝关节OA患者中，预测的MCF峰值与内侧胫骨软骨体积损失呈正相关。
Objective. The relationship between in vivo knee load predictions and longitudinal cartilage changes has not been investigated. We undertook this study to develop an equation to predict the medial tibiofemoral contact force (MCF) peak during walking in persons with instrumented knee implants, and to apply this equation to determine the relationship between the predicted MCF peak and cartilage loss in patients with knee osteoarthritis (OA).Methods. In adults with knee OA (39 women, 8 men; mean +/- SD age 61.1 +/- 6.8 years), baseline biomechanical gait analyses were performed, and annualized change in medial tibial cartilage volume (mm(3)/year) over 2.5 years was determined using magnetic resonance imaging. In a separate sample of patients with force-measuring tibial prostheses (3 women, 6 men; mean +/- SD age 70.3 +/- 5.2 years), gait data plus in vivo knee loads were used to develop an equation to predict the MCF peak using machine learning. This equation was then applied to the knee OA group, and the relationship between the predicted MCF peak and annualized cartilage volume change was determined.Results. The MCF peak was best predicted using gait speed, the knee adduction moment peak, and the vertical knee reaction force peak (root mean square error 132.88N; R-2 = 0.81, P < 0.001). In participants with knee OA, the predicted MCF peak was related to cartilage volume change (R-2 = 0.35, beta = -0.119, P < 0.001).Conclusion. Machine learning was used to develop a novel equation for predicting the MCF peak from external biomechanical parameters. The predicted MCF peak was positively related to medial tibial cartilage volume loss in patients with knee OA.