Joint Track Machine Learning: An Autonomous Method of Measuring Total Knee Arthroplasty Kinematics From Single-Plane X-Ray Images.

BACKGROUND: Dynamic radiographic measurements of 3-dimensional (3-D) total knee arthroplasty (TKA) kinematics have provided important information for implant design and surgical technique for over 30 years. However, current methods of measuring TKA kinematics are too cumbersome, inaccurate, or time-consuming for practical clinical application. Even state-of-the-art techniques require human-supervision to obtain clinically reliable kinematics. Eliminating human supervision could potentially make this technology practical for clinical use. METHODS: We demonstrate a fully autonomous pipeline for quantifying 3D-TKA kinematics from single-plane radiographic imaging. First, a convolutional neural network (CNN) segmented the femoral and tibial implants from the image. Second, those segmented images were compared to precomputed shape libraries for initial pose estimates. Lastly, a numerical optimization routine aligned 3D implant contours and fluoroscopic images to obtain the final implant poses. RESULTS: The autonomous technique reliably produces kinematic measurements comparable to human-supervised measures, with root-mean-squared differences of less than 0.7 mm and 4° for our test data, and 0.8 mm and 1.7° for external validation studies. CONCLUSION: A fully autonomous method to measure 3D-TKA kinematics from single-plane radiographic images produces results equivalent to a human-supervised method, and may soon make it practical to perform these measurements in a clinical setting.

Automated Registration of 3-D Knee Implant Models to Fluoroscopic Images Using Lipschitzian Optimization.

This paper describes an automated method for registering 3-D models of metallic knee implants to single-plane radiographic images. We develop a multistage approach that identifies the correct pose by matching altered dilations of an edge-detected image with the silhouette of an implant model. The location of the similarity function's minimum is found using a novel optimization routine that combines the Dividing Rectangles algorithm with properties of the registration metric. Depending on the implant type (tibial or femoral), this technique reliably converges under maximum displacements of approximately 25 to 55 millimeters for translation components and 25° to 55° for Euler angles. The method proves to be robust to noise from bones and soft tissue. After an initial guess for the first image in the sequence, subsequent frames can be automatically registered from the optimum pose in the previous image.

Automated Registration of Three-Dimensional Knee Implant Models to Fluoroscopic Images using Lipschitzian Optimization.

This paper describes an automated method for registering three-dimensional models of metallic knee implants to single-plane radiographic images. We develop a pyramidal approach that identifies the correct pose by matching decreasing dilations of an edge-detected image with the silhouette of an implant model. The location of the similarity function's minimum is found using a novel optimization routine that combines the DIRECT (Dividing Rectangles) algorithm with properties of the Lipschitz constant specific to the registration metric. Depending on the implant type, this technique reliably converges under maximum displacements of approximately 25 to 55 millimeters for translation components and 25 to 55 degrees for Euler angles. The method proves to be robust to noise from bones and soft tissue. After an initial guess for the first image in the sequence, subsequent frames can be automatically registered from the optimum pose in the previous image. Once optimized, the poses from the better-performing femoral implants can be used to create an image mask and slightly increase tibial registration success while maintaining automation.