Feasibility of using MRIs to create subject-specific parallel-mechanism joint models.

Musculoskeletal models typically use generic 2D models for the tibiofemoral (TFJ) and patellofemoral (PFJ) joints, with a hinge talocrural joint (TCJ), which are scaled to each subject׳s bone dimensions. Alternatively joints' measured kinematics in cadavers are well-predicted using 3D cadaver-specific models. These employ mechanisms constrained by the articulations of geometric objects fitted to the joint׳s surfaces. In this study, we developed TFJ, PFJ and TCJ mechanism-based models off MRIs for fourteen participants and compared the estimated kinematics with those from published studies modified to be consistent with mechanisms models and subject-specific anatomical landmarks. The models' parameters were estimated by fitting spheres to segmented articular cartilage surfaces, while ligament attachment points were selected from their bony attachment regions. Each participant׳s kinematics were estimated by ensuring no length changes in ligaments and constant distances between spheres' centres. Two parameters' optimizations were performed; both avoid singularities and one best matches the kinematic patterns off published studies. Sensitivity analysis determined which parameters the models were sensitive to. With both optimization methods, kinematics did not present singularities but correlation values were higher, exceeding 0.6, when matching the published studies. However, ranges of motion (ROM) were different between estimated and published studies. Across participants, models presented large parameter variation. Small variations were found between estimated- and optimized-parameters, and in the estimated-rotations and translations' means and ROM. Model results were sensitive to changes in distal tibia, talus and patella spheres' centres. These models can be implemented in subject-specific rigid-body musculoskeletal models to estimate joint moments and loads.