Wearable-based estimation of continuous 3D knee moments during running using a convolutional neural network.

IF 2 3区医学 Q3 ENGINEERING, BIOMEDICAL

Sports Biomechanics Pub Date : 2025-03-24 DOI:10.1080/14763141.2025.2481164

Lucas Höschler, Christina Halmich, Christoph Schranz, Julian Fritz, Saša Čigoja, Martin Ullrich, Anne D Koelewijn, Hermann Schwameder

{"title":"Wearable-based estimation of continuous 3D knee moments during running using a convolutional neural network.","authors":"Lucas Höschler, Christina Halmich, Christoph Schranz, Julian Fritz, Saša Čigoja, Martin Ullrich, Anne D Koelewijn, Hermann Schwameder","doi":"10.1080/14763141.2025.2481164","DOIUrl":null,"url":null,"abstract":"<p><p>This study aimed to develop and validate a machine learning method to estimate continuous 3D knee moments during running from wearable sensor data. Reference knee moments were calculated from 19 recreational runners during treadmill running at varying slopes (0 ± 5 % incline), speeds (self-selected ± 1 km/h) and in 3 types of footwear. A convolutional neural network was trained on data from 7 inertial measuring units (feet, shanks, thighs, sacrum) and a pair of pressure insoles. We assessed performance over continuous time windows (CONT) and during stance phases (PHSS) by intraclass-correlation (ICC), normalised root mean squared error (nRMSE), and statistical parametric mapping. The agreement levels in the sagittal plane were good to excellent (ICC: 0.84-0.98), with low errors (nRMSE: 0.05-0.11). However, accuracy was lower for non-sagittal estimations (frontal ICC: 0.19-0.90, nRMSE: 0.08-0.23; transverse ICC: 0.72-0.94, nRMSE: 0.07-0.17). Accuracy decreased across all planes during PHSS. The proposed approach yields similar or better accuracy compared to previous work while requiring less preprocessing. It provides a viable method for wearable-based assessment of running kinetics in near real-time. Additional data and methods to address inter-individual variability could improve its precision in assessing frontal plane injury risk factors.</p>","PeriodicalId":49482,"journal":{"name":"Sports Biomechanics","volume":" ","pages":"1-19"},"PeriodicalIF":2.0000,"publicationDate":"2025-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Sports Biomechanics","FirstCategoryId":"5","ListUrlMain":"https://doi.org/10.1080/14763141.2025.2481164","RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ENGINEERING, BIOMEDICAL","Score":null,"Total":0}

引用次数: 0

Abstract

This study aimed to develop and validate a machine learning method to estimate continuous 3D knee moments during running from wearable sensor data. Reference knee moments were calculated from 19 recreational runners during treadmill running at varying slopes (0 ± 5 % incline), speeds (self-selected ± 1 km/h) and in 3 types of footwear. A convolutional neural network was trained on data from 7 inertial measuring units (feet, shanks, thighs, sacrum) and a pair of pressure insoles. We assessed performance over continuous time windows (CONT) and during stance phases (PHSS) by intraclass-correlation (ICC), normalised root mean squared error (nRMSE), and statistical parametric mapping. The agreement levels in the sagittal plane were good to excellent (ICC: 0.84-0.98), with low errors (nRMSE: 0.05-0.11). However, accuracy was lower for non-sagittal estimations (frontal ICC: 0.19-0.90, nRMSE: 0.08-0.23; transverse ICC: 0.72-0.94, nRMSE: 0.07-0.17). Accuracy decreased across all planes during PHSS. The proposed approach yields similar or better accuracy compared to previous work while requiring less preprocessing. It provides a viable method for wearable-based assessment of running kinetics in near real-time. Additional data and methods to address inter-individual variability could improve its precision in assessing frontal plane injury risk factors.

查看原文本刊更多论文

基于可穿戴设备的跑步过程中连续三维膝关节力矩的卷积神经网络估计。

本研究旨在开发和验证一种机器学习方法，从可穿戴传感器数据中估计跑步过程中连续的3D膝关节力矩。19名休闲跑步者在不同坡度（0±5%倾斜度）、速度（自行选择±1 km/h）和3种鞋类的跑步机上跑步时，计算了参考膝关节力矩。卷积神经网络对来自7个惯性测量单元（脚、小腿、大腿、骶骨）和一双压力鞋垫的数据进行训练。我们通过类内相关（ICC）、归一化均方根误差（nRMSE）和统计参数映射来评估连续时间窗（CONT）和姿态阶段（PHSS）的性能。矢状面一致性水平为良至优（ICC: 0.84-0.98），误差较低（nRMSE: 0.05-0.11）。然而，非矢状面估计的准确性较低(额部ICC: 0.19-0.90, nRMSE: 0.08-0.23；横向ICC: 0.72-0.94, nRMSE: 0.07-0.17)。在PHSS期间，所有平面的精度都下降了。与以前的工作相比，所提出的方法产生相似或更好的精度，同时需要较少的预处理。它为基于可穿戴设备的近实时运行动力学评估提供了一种可行的方法。更多的数据和方法来解决个体间的差异，可以提高其评估额叶损伤危险因素的准确性。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Sports Biomechanics 医学-工程：生物医学

CiteScore

5.70

自引率

9.10%

发文量

135

审稿时长

>12 weeks

期刊介绍： Sports Biomechanics is the Thomson Reuters listed scientific journal of the International Society of Biomechanics in Sports (ISBS). The journal sets out to generate knowledge to improve human performance and reduce the incidence of injury, and to communicate this knowledge to scientists, coaches, clinicians, teachers, and participants. The target performance realms include not only the conventional areas of sports and exercise, but also fundamental motor skills and other highly specialized human movements such as dance (both sport and artistic). Sports Biomechanics is unique in its emphasis on a broad biomechanical spectrum of human performance including, but not limited to, technique, skill acquisition, training, strength and conditioning, exercise, coaching, teaching, equipment, modeling and simulation, measurement, and injury prevention and rehabilitation. As well as maintaining scientific rigour, there is a strong editorial emphasis on ''reader friendliness''. By emphasising the practical implications and applications of research, the journal seeks to benefit practitioners directly. Sports Biomechanics publishes papers in four sections: Original Research, Reviews, Teaching, and Methods and Theoretical Perspectives.