Orthopaedic disorders are a leading cause of disability in the U.S., with arthritis and/or spine problems adversely affecting quality of life fo more than 20% of adults. With an aging population the rate of disability from orthopaedic disorders has been increasing steadily. While advances in diagnostic imaging (including CT, MRI and ultrasound) have greatly improved our ability to detect structural changes in musculoskeletal tissues, they typically reveal little about joint function. There is evidence that abnormal mechanical joint function contributes significantly to the development and progression of many types of joint disease. There is, therefore, a significant clinical need for the widespread use of technologies that can identify subtle abnormalities in joint function that, if left untreate, can compromise long-term joint health. Biomechanical analyses are a key tool for providing quantitative objective measures of patient status and treatment outcomes. At the heart of most in vivo biomechanical analyses is the estimation of the position and orientation (Pose) of a multi-segment rigid body model based on recordings of 3D motion sensor data. The principal assumption of existing Pose estimation algorithms is that the motion sensors move rigidly along with the body segments to which they are attached; it is known, however, that this assumption is an approximation and that the sensors in reality move relative to the underlying skeleton. This project is designed to apply algorithms, based on Bayesian Inference, which have the potential for mitigating soft tissue artifacts and dramatically improving the spatial resolution of 3D movement analysis. To address this soft-tissue problem we redefined Pose estimation using the general framework of probabilistic (Bayesian) inference. In Phase I we developed a general Bayesian Prior based on soft tissue motion that produced substantially lower errors than all generative methods. In Phase II a new Bayesian Priors will be implemented to mitigate soft tissue artifact based on DSX data of the knee and ankle for a set of control subjects. In Phase II we will test this Probabilistic Inference approach against an independent set normal subjects during walking and running, against a set of subjects with Cerebral Palsy during walking, and against a set of subjects with Anterior Collateral Ligament (ACL) injuries during walking and running. The improvement in spatial resolution demonstrated in Phase I and the enhancements proposed for Phase II will enable non-invasive Motion Capture to achieve sufficiently high spatial accuracy to describe the dynamic functioning of joints and ligaments, which will lead to an experimental and analytical tool suitable for studying joint disease and disorders.
Biomechanical analyses based on 3D Motion Capture are a key tool for establishing quantitative, objective measures of functional movement status and treatment outcomes. In Phase I we implemented a probabilistic (Bayesian) approach for estimating the position and orientation of anatomical bodies and found that when tested against Dynamic Stereo X-ray data the approach was more accurate than existing discriminative solutions and demonstrated the potential to achieve sufficiently high spatial accuracy to study joint disease and disorders.
