Disorders of human joints manifest during dynamic movement, yet no objective tools are widely available for clinicians to assess or diagnose abnormal joint motion during functional activity. Machine learning tools have supported advances in many applications for image interpretation and understanding and have the potential to enable clinically and economically practical methods for objective assessment of human joint mechanics. We performed a study using convolutional neural networks to autonomously segment radiographic images of knee replacements and to determine the potential for autonomous measurement of knee kinematics. The autonomously segmented images provided superior kinematic measurements for both femur and tibia implant components. We believe this is an encouraging first step towards realization of a completely autonomous capability to accurately quantify dynamic joint motion using a clinically and economically practical methodology.

Model-image registration types of measurements have profoundly changed capabilities for studying dynamic 3D joint and implant kinematics since their introduction in the early 1990's. Since that time, a variety of proprietary and open-source software packages have been developed and reported for performing these measurements. Model-image registration based measurements have been used to quantify motions in natural and replaced knees, hips, ankles, shoulders, elbows, and spines in both single- and stereo-projection radiographic measurement setups. In theory, with the same quality images and the same quality bone/implant models, any of the software developed to perform model-image registration has the potential to provide equivalent measurement accuracy. Hence, much of the effort to improve measurement capabilities has been to reduce human interaction requirements and make the measurements more automatic and objective. In this paper, we report a new open-source software program that requires a minimum of user input to automate the 3D kinematic measurement process from single- or bi-plane radiographic projections.

JointTrack Auto (JTA) is an open source (sourforge.net/projects/jointtrackauto) program for performing model-image registration of metallic implants with single- or bi-plane radiographic images (see image). A predominantly edge-based cost function is used with an adaptive partioning global optimization scheme for model-image registration. Although this method works without any human intervention, JTA allows users to roughly identify one ‘feature’ of each implant that is visible in all images, e.g. the tip of a peg, to very significantly reduce the search space and time required for numerical optimization. This makes for a very convenient and fast initialization process where a human user simply mouse-clicks on a few easily identifiable locations in each radiographic image, and then the automated registration process is begun.

Registration accuracy examples and a software demonstration will be included in this e-poster presentation to introduce attendees to the software and spur discussion about the various methods available to perform these important measurements.