Assessing the Validity of Wearable Inertial Sensors in Evaluating Joint Kinetics and Hamstring Musculotendon Mechanics at Various Running Speeds

ABSTRACT Purpose Integrating musculoskeletal (MSK) modeling with inertial measurement units (IMUs) offers a promising approach for analyzing joint and muscle function during locomotion. This study examined the validity of combining IMUs, MSK modeling, and inverse dynamics to estimate lower-limb joint moments and hamstring musculotendon (MT) mechanics during treadmill running at varying speeds. Methods Eighteen healthy young adults ran on a treadmill at 70% (5.21 ± 0.62 m·s −1 ), 80% (5.96 ± 0.71 m·s −1 ), 85% (6.33 ± 0.76 m·s −1 ), 90% (6.70 ± 0.80 m·s −1 ), 95% (7.07 ± 0.84 m·s −1 ), and 100% (7.44 ± 0.89 m·s −1 ) of their maximal sprinting speed. Kinematic data were simultaneously collected using both an optical motion capture (OMC) system (Vicon) and an IMU system (Xsens), whereas electromyographic data recorded hamstring activity. MSK modeling was applied to both kinematic measurements to calculate lower-limb joint moments and hamstring MT mechanics, with estimated muscle activations validated against the electromyographic data. Results IMU-based estimations closely matched OMC-based calculations, with coefficient of multiple correlations exceeding 0.85 for hip and knee joint moments during swing and 0.95 for hamstring MT kinematics across full stride cycles at all speeds. MT force estimations varied among hamstring muscles, with semimembranosus showing the highest agreement (0.96 < coefficient of multiple correlation < 0.98) across all speeds. Linear mixed models showed for each 1 m·s −1 speed increase, root mean square errors between the two systems increased by less than 0.25 N·m for joint moments and 0.05 body weight for hamstring MT forces. Conclusions IMU-MSK integration is a valid alternative to OMC for estimating sagittal-plane joint moments and hamstring MT mechanics during treadmill running, although differences in peak hip moment during terminal swing warrant caution in field-based applications.