A multi method (MSK Model vs Geometry Based vs BVR) comparison and bone-model (Generic scaled vs MRI derived) sensitivity analysis on estimates of knee joint kinematics.

Accurate 3D knee kinematics are essential for building reliable musculoskeletal (MSK) models. This is especially true when studying complex internal mechanisms like joint contact forces or the progression of degenerative pathologies. Traditionally, simulation environments like OpenSim rely on motion capture (MoCap) and force plate data to estimate these kinematics, constraining joint motion to an Anatomical Coordinate System (ACS) predefined within a scaled generic model. While accepted as a standard, this approach is highly sensitive to soft tissue artifacts and depends heavily on the initial definition of the joint axes. The human knee exhibits highly coupled motions across all six degrees of freedom. Therefore, even minor misalignments in the ACS can introduce significant kinematic cross-talk. On the other hand, experimental methods like biplane videoradiography (BVR) offer high-fidelity spatial kinematics by applying geometric fitting algorithms directly to patient-specific bone models. Due to fundamental differences in defining the ACS, directly comparing MSK model-based estimates to BVR remains a non-trivial task. This project addresses this gap through a multi-method comparative framework. We evaluate the 3D knee kinematics generated by a state-of-the-art MSK optimization algorithm (COMAK) against those calculated using a strictly geometry-based approach, utilizing in vivo BVR data as the ground truth for validation. Concurrently, we conduct a sensitivity analysis on the bone models to isolate the effect of anatomical personalization, comparing generic scaled geometries against MRI-derived, subject-specific meshes. To enable this comparison, the core technical task involves a custom computational pipeline that extracts dynamic 3D bone geometries (femur and tibia) at each discrete time step of the simulated movement. By replicating the mathematical logic of BVR, the algorithm performs an automated geometric fit on these time-series surfaces to define a consistent, recalculable ACS. Ultimately, by comparing the generic and personalized kinematic trajectories with the experimental gold standard, this work provides a rigorous quantification of how sensitive kinematic estimations are to both the tracking methodology and the anatomical fidelity of the underlying bone models, offering practical insights for the future of personalized knee biomechanics. Finally, we will run a comparative analysis across the different methodologies and bone models. By evaluating both generic scaled and MRI-derived geometries against the biplanar videoradiography (BVR) gold standard, we can quantify how sensitive kinematic estimations are to both the underlying tracking methodology and subject-specific bone modeling. Ultimately, this work will provide practical insights into the limitations and potential of coordinate system definitions and personalized anatomical modeling in knee biomechanics.