OpenOFM: an open-source implementation of the multi-segment Oxford Foot Model

Philippe Dixon, Elodie Drew, Sean McBride, Samuel Cheng, Marian Harrington,Julie Stebbins,Amy Zavatsky

Gait & Posture(2023)

引用 0|浏览3
暂无评分
摘要
Reproducibility is crucial in biomechanics research and relies heavily on transparency in computational approaches [1]. Biomechanical models are especially vulnerable to a lack of methodological detail, leading to limited reproducibility of results [2]. These issues are of particular concern in multi-segment foot modelling, given the complexity of methods used to extract three-dimensional intra-foot segment kinematics from skin-mounted markers. The Oxford Foot Model (OFM) is one of the most used multi-segment foot models, and its clinical relevance has been demonstrated through multiple studies [5]. Unfortunately, the “black-box” implementation of the model developed by Stebbins et al. [6] in Vicon’s Nexus Software (Oxford Metrics, Inc., United-Kingdom) prevents code inspection. The aim of this project is to develop an open-source version of the OFM (openOFM) coded in a high-level programming language to improve accessibility and interpretation of the model. The OFM estimates five angles in three planes, using four segments: the tibia, hindfoot, forefoot, and hallux, defined by a total of 43 markers, with 13 on each foot. The accuracy of our open-source model is evaluated by comparing our kinematic results to those generated by the Vicon OFM plug-in (version 2.13) using the root mean squared error, normalized to the magnitude of each angle (NRMSE). Three trials from participants with distinct foot morphologies (normal, pes cavus, pes planus) were selected from our gait database. Trials were collected using a 6-camera system at 100 Hz. Average kinematics for the gait cycle were computed for each angle, with and without the use of foot flat options (forefoot or hindfoot flat). The openOFM code was developed in Matlab (2022a, The Mathworks, Inc., Natick, USA). There is strong agreement between the two OFM versions. On average, the NRMSE between the models is of 0.002, irrespective of the options used. The average curves obtained using both models for each angle on the left side are shown in Fig. 1.Download : Download high-res image (125KB)Download : Download full-size image Fig. 1: Average tibia and foot angle estimations (degrees) of the three trials (one trial per foot morphology) in the sagittal (a–e), coronal (f-i), and transverse (j-n)plane via openOFM and the Vicon OFM plug-in (Vicon OFM) for the left limb without using the foot flat options (normalized to 100% of the gait cycle). Our accurate, open-source reproduction of the OFM will allow users to freely implement the model and inspect the code to gain confidence in the methods and increase understanding of results. We are now assessing the robustness of the code for use in patients with severe foot pathologies and developing a fully open-access version of the code in Python. Subsequently, the model will be updated to better reflect underlying foot motion. A public code repository will be released on GitHub.
更多
查看译文
关键词
foot,model,open-source,multi-segment
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要