Chair of Mechanics and Robotics, University of Duisburg-Essen, Lotharstraße 1, 47057, Duisburg, Germany.
Chair of Orthopaedics and Trauma Surgery, University of Duisburg-Essen, Essen, Germany.
Int Orthop. 2023 Apr;47(4):921-928. doi: 10.1007/s00264-022-05670-0. Epub 2023 Jan 10.
Orthopaedic scores are essential for the clinical assessment of movement disorders but require an experienced clinician for the manual scoring. Wearable systems are taking root in the medical field and offer a possibility for the convenient collection of motion tracking data. The purpose of this work is to demonstrate the feasibility of automated orthopaedic scorings based on motion tracking data using the Harris Hip Score and the Knee Society Score as examples.
Seventy-eight patients received a clinical examination and an instrumental gait analysis after hip or knee arthroplasty. Seven hundred forty-four gait features were extracted from each patient's representative gait cycle. For each score, a hierarchical multiple regression analysis was conducted with a subsequent tenfold cross-validation. A data split of 70%/30% was applied for training/testing.
Both scores can be reproduced with excellent coefficients of determination R for training, testing and cross-validation by applying regression models based on four to six features from instrumental gait analysis as well as the patient-reported parameter 'pain' as an offset factor.
Computing established orthopaedic scores based on motion tracking data yields an automated evaluation of a joint function at the hip and knee which is suitable for direct clinical interpretation. In combination with novel technologies for wearable data collection, these computations can support healthcare staff with objective and telemedical applicable scorings for a large number of patients without the need for trained clinicians.
矫形评分对于运动障碍的临床评估至关重要，但需要经验丰富的临床医生进行手动评分。可穿戴系统正在医学领域扎根，并为方便地收集运动跟踪数据提供了一种可能性。本研究旨在展示基于运动跟踪数据使用 Harris Hip Score 和 Knee Society Score 进行自动矫形评分的可行性。
78 名接受髋关节或膝关节置换术的患者接受了临床检查和仪器步态分析。从每位患者的代表性步态周期中提取了 744 个步态特征。对于每个评分，使用分层多元回归分析，并随后进行了十折交叉验证。应用 70%/30%的数据分割进行训练/测试。
通过应用基于仪器步态分析的四个至六个特征以及患者报告的“疼痛”参数作为偏移因子的回归模型，两种评分在训练、测试和交叉验证中都可以通过优秀的决定系数 R 进行再现。
基于运动跟踪数据计算既定的矫形评分可对髋关节和膝关节的关节功能进行自动评估，适合直接临床解释。结合可穿戴数据采集的新技术，这些计算可以为大量患者提供客观且适用于远程医疗的评分，而无需经过培训的临床医生。
