Novel Exposure Monitoring for FAIS Prevention in Sport: Identifying High-risk Hip Motions in Soccer
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Waterloo
Abstract
Hip osteoarthritis (OA) is a leading cause of pain and disability worldwide and is more prevalent among former elite athletes, particularly those who participated in high-impact sports during adolescence. Femoroacetabular impingement syndrome (FAIS), a common cause of hip and groin pain in young, physically active adults, is increasingly recognized as an important precursor to early-onset hip OA. Participation in high-impact youth sports, particularly soccer, has been strongly associated with the development of cam morphology, a bony protrusion of the femoral head–neck junction, and subsequent symptomatic cam FAIS. However, the sport-specific hip kinematics and movement exposures thought to contribute to this condition remain poorly understood. While repetitive exposure to combined hip flexion, adduction, and internal rotation is thought to contribute to symptomatic cam FAIS, no validated tools currently exist to quantify the magnitude and frequency of exposure to these provocative hip positions during ecologically valid sport participation. This thesis contributed to addressing this gap through two complementary objectives. First, a wearable inertial measurement unit (IMU) system was validated against a gold-standard optical motion capture (OMC) system for quantifying 3D hip kinematics, encompassing sagittal, frontal, and transverse plane motion, during 35 soccer-specific movements performed by 20 elite adult soccer players (10 male, 10 female). Second, five supervised machine-learning classifiers were developed and evaluated to automatically identify a subset of 20 soccer-specific movements from IMU-derived signals. These movements were selected based on their kinematic relevance to cam FAIS and ecological relevance to soccer participation.
The IMU system demonstrated no significant differences between the right and left hip joint angles in sagittal (mean difference = -0.60˚ to -1.01˚, p = 0.171) or frontal (mean difference = -0.32˚ to +0.13˚, p = 0.223) plane range of motion, with very good-to-excellent waveform agreement with OMC, as indicated by the coefficient of multiple correlation (CMC2 = 0.964–0.967 and 0.853–0.876, respectively). Transverse plane internal/external rotation demonstrated a significant underestimation of ROM by the IMU (mean difference = -9.09˚ to -12.28˚, p < 0.001) and lower waveform agreement with OMC (CMC2 = 0.750–0.761) compared with the sagittal and frontal planes. Among the five ML classifiers evaluated, XGBoost achieved the highest overall classification performance, accurately distinguishing 20 soccer-specific movements with a macro-averaged F1-score of 0.905 and overall accuracy of 94.4%. When evaluated on continuous sliding-window data from unseen participants, classification accuracy decreased to 72.0% (F1-score = 0.759), although misclassifications remained concentrated between kinematically similar technical actions rather than distinct movement categories.
Collectively, these findings demonstrate the feasibility of combining wearable IMUs with machine learning to provide estimates of sagittal and frontal plane hip joint angles and to distinguish a comprehensive range of soccer-specific movements associated with hip positions implicated in symptomatic cam FAIS. While further work is required to improve transverse plane measurement accuracy, develop classifiers using continuous movement data, and validate the framework in adolescent populations, this research establishes an important foundation for quantifying real-world exposure to sport-specific hip kinematics. This approach may extend beyond soccer to support movement exposure monitoring across a broader range of athletic populations at risk of cam FAIS, ultimately contributing not only to characterizing cam FAIS-related movement risk, but to the broader goal of reducing the burden of hip OA in physically active populations.