Wearable Estimation of Knee Joint Contact Forces: SiMPLeSENSE, a Simulated-IMU Pipeline Validated Against In Vivo Loads

Abstract

Knee joint compressive forces (KCF) are central to understanding joint loading and degeneration, but they can be measured directly only with instrumented implants. Musculoskeletal models provide a non-invasive alternative, and inertial measurement units (IMUs) could make such estimates portable. However, no dataset combines IMU signals with in vivo implant forces, so the accuracy of IMU-based KCF remains unverified. We developed SiMPLeSENSE, a framework that derives virtual IMU orientations from optical motion capture using subject-specific SMPL meshes. Using the CAMS-Knee dataset, we reconstructed kinematics with OpenSense and estimated KCF in OpenSim with a Rapid Muscle Redundancy solver, comparing the estimates with in vivo measurements during level walking and stair descent. We also assessed robustness to sensor placement error by adding empirically derived thigh and shank orientation noise (for two cases). Across subjects and activities, the simulated-IMU estimates closely tracked the optical mocap-driven estimates and reproduced the in vivo force pattern, with the agreement against the optical reference being higher for stair descent. Overall, the estimated forces matched the in vivo measurements with a median RMSE of about 0.3 to 0.5 BW and correlations near 0.95. Adding realistic orientation error increased RMSE only modestly, whereas doubling it markedly overestimated peak force, most evident for stair descent. These results indicate that the approach is robust to substantial changes in input kinematics, supporting a laboratory-independent setup combining inertial motion capture, a small EMG set, and instrumented force shoes for home and clinical monitoring of knee loading.