Assessing Single and Dual-Sensor IMU Setup for 3D Foot Modelling in Running

Background: Accurate measurement of foot kinematics is essential for gait analysis, injury prevention, and performance assessment. While optical motion capture (OMC) remains the gold standard for multi-segment foot modelling, inertial measurement units (IMUs) offer a portable and inexpensive alternative. However, most IMU-based approaches rely on a single foot-mounted sensor, treating the foot as one rigid segment and neglecting independent foot segment motion. Objective: This study investigates the accuracy of a dual-sensor IMU setup compared to a single-sensor IMU setup for estimating hindfoot-forefoot (HF/FF) joint angles using a multi-segment foot model during running. Accuracy is evaluated relative to optical motion capture (OMC) as the reference standard and with additional focus on the influence of running speed and foot strike pattern, forefoot striker (FFS) vs. rearfoot striker (RFS). Methods: Six healthy recreational runners (3 rearfoot strikers, 3 forefoot strikers) ran on an instrumented treadmill at two speeds (9 and 11 kph), while foot segment orientations were recorded using inertial measurement units (IMUs) and optical motion capture (OMC). A dual-IMU setup captured hindfoot and forefoot orientation directly, while a single-IMU setup estimated forefoot orientation from the hindfoot. HF/FF joint angles from both IMU setups were compared to OMC using root mean square error (RMSE) and standard deviation (SD) across the gait cycle, both overall and separately by strike pattern. Additionally, per-subject error mean and SD, as well as correlation values between the IMU setups and the OMC, were calculated. Results: Both IMU setups show similar overall error patterns, with the largest differences around toe-off. The dual-IMU setup showed improved accuracy during push-off, but slightly higher variability across the other phases. RMSE and SD increased at the higher speed (11 kph). Correlation with OMC was generally higher for the dual-IMU setup, yet no consistent differences in mean error were observed between the two IMU configurations. While running speed affected both setups similarly, by increasing error and lowering correlation, the strike pattern had minimal effect on performance. Conclusion: Given the added complexity of dual-sensor configurations, single-IMU setups may be sufficient for general gait analysis, whereas the dual-IMU setup may be preferable for applications requiring detailed forefoot motion tracking or improved accuracy during dynamic push-off phases.