Human Kinematics Reconstruction from Markerless and Marker-based Motion Analysis Systems by Model-based Approach

INTRODUCTION Clinically relevant muscle information (length, moment arms) is difficult to estimate directly in clinical settings. Quality of the obtained muscle approximation strongly relies on the underlying bone and kinematic data used to create the joint models that will be crossed by the muscle spatial path. We present a method that allows fusing accurate joint kinematic information with relatively crude motion analysis (MA) data collected using either marker-based stereophotogrammetry (i.e. bone displacement collected from reflective markers glued on the subject's skin) or markerless single-camera hardware. The obtained kinematical model can then be used for further modeling (e.g. muscle moment arm or excursion by addition of relevant data) which quality will depend on the underlying kinematic model. Typically, a global optimization method based on mechanical modeling could be applied to adjust model parameters to a particular motion. Different sets of joint constraints related to joint kinematics (e.g. joint surface geometry, ligament information and joint mechanism) were previously implemented in order to assess their influence on the lower limb kinematics during gait [1]. This approach requests implementation of collision detection and reaction mechanism procedures such as the ones available from most commercial multibody dynamics software.