To achieve 3D kinematic analysis of artificial knee implants, 2D/3D registration technique, which uses X-ray fluoroscopic images and computer assisted design model of the knee implant, has been applied to clinical cases. In previous method, however, we had to set the initial pose of 2D/3D registration with manual operations, and that was one of the problems for practical clinical applications. In this paper, we propose an improved method to automatically determine the initial pose using rigid body tracking with feature points of X-ray images. As a result of experiments using actual fluoroscopic images, the success rate with proposed method is higher than conventional method. In addition, for femoral component, average error with conventional and proposed method was 3.93mm/2.61°, 0.79mm/1.93°, respectively. For tibial component, average error with conventional and proposed method was 3.56mm/1.03°, 0.89mm/0.81°, respectively. We consider the proposed method would be more effective for initial pose estimation than the conventional method.
Published in:
Computer Sciences and Convergence Information Technology (ICCIT), 2011 6th International Conference on
Date of Conference: Nov. 29 2011-Dec. 1 2011