Fuzzy Image Matching for Pose Recognition of Occluded Knee Implants Using Fluoroscopy Images

Abstract

Fluoroscopy images have been widely used for evaluating kinematics of the knee implant in vivo after the total knee Arthroplasty, TKA in short. The knee implant mainly consists of tibial tray, tibial insert and femoral component. Because a fluoroscopy image is a 2-D projection image of the 3-D knee implant, the tibial tray often overlaps with the femoral component on the projection image, or a part of the knee implant is outside of the field of view (FOV). In order to analyze such occluded images, this article introduces fuzzy logic into image matching of the given 2-D fluoroscopy image and 3-D geometric models of the knee implant. Based on the proposed fuzzy image matching algorithm, we present a novel computer-aided diagnosis (CAD) system for estimating 3-D kinematics of the knee implant with 2-D fluoroscopy dynamic images. To quantitatively evaluate our system, it was applied to computer-simulated images and phantom images that took the knee implant in vitro fixed with arbitrary pose by a jig. The experimental results denoted that this system could estimate the pose of the knee implant within the error of 0.86° with non-occluded images, and within the error of 1.28° with 15% occluded images. Also, the proposed system was applied to two patients after TKA to demonstrate the clinical application.