Accuracy of feature-based 2D/3D registration for analyzing normal knee kinematics using X-ray fluoroscopy and CT images

摘要

This study presents accuracy of feature-based 2D/3D registration for analyzing normal knee kinematics using X-ray fluoroscopy with FPD and CT images. Kinematics of normal knee is determined using a 2D/3D registration technique based on knee silhouette contour, which uses 3D bone polygonal surface model to estimate the 3D pose of femur, tibia/fibula, and patella from a dynamic FPD images. In order to validate the pose estimation accuracy of normal knee, computer simulation test and phantom experiments were performed. The result of computer simulation test showed that the root mean square errors were within 1.0 mm and 1.0 degree except for out-of-plane translation for femur and tibia/fibula. While, for patella, rotation errors were increased. The result of phantom experiments also showed similar trends as observed for computer simulation test, and total errors were increased. Consequently, a feature-based 2D/3D registration for the kinematic analysis of normal knee was performed successfully for femur and tibia/fibula, but patella analysis required further investigation.