Abstract
Purpose
To achieve 3D kinematic analysis of total knee arthroplasty (TKA), 2D/3D registration techniques, which use X-ray fluoroscopic images and computer aided design model of the knee implants, have been applied to clinical cases. However, most conventional methods have needed time-consuming and labor-intensive manual operations in some process. In particular, for the 3D pose estimation of tibial component model from X-ray images, these manual operations were carefully performed because the pose estimation of symmetrical tibial component get severe local minima rather than that of unsymmetrical femoral component. In this study, therefore, we propose an automated 3D kinematic estimation method of tibial component based on statistical motion model, which is created from previous analyzed 3D kinematic data of TKA.
Methods
The used 2D/3D registration technique is based on a robust feature-based (contour-based) algorithm. In our proposed method, a statistical motion model which represents average and variability of joint motion is incorporated into the robust feature-based algorithm, particularly for the pose estimation of tibial component. The statistical motion model is created from previous a lot of analyzed 3D kinematic data of TKA. In this study, a statistical motion model for relative knee motion of the tibial component with respect to the femoral component was created and utilized. Fig. 1 shows each relative knee motion model for six degree of freedom (three translations and three rotations parameter). Thus, after the pose estimation of the femoral component model, 3D pose of the tibial component model is determined by maximum a posteriori (MAP) estimation using the new cost function introduced the statistical motion model.
Experimental results
To validate the feasibility and effectiveness of 3D pose estimation for the tibial component using the proposed method, experiments using X-ray fluoroscopic images of 20 TKA patients under the squatting knee motion were performed. For the creation of correct pose (reference data) and the statistical motion model, we used the 3D pose data which were got by carefully applying previous method to the contour images which spurious edges and noises were removed manually. In order to ensure the validity for the statistical motion model of the proposed method, leave-one-out cross validation method was applied. In the 3D pose estimation of tibial component model, for the only first frame, initial guess pose of the model was manually given. For all images except for the first frame, the 3D pose of the model was automatically estimated without manual initial guess pose of the model. To assess the automation performance, the automation rate was calculated, and the rate was defined as the X-ray frame number of satisfying clinical required accuracy (error within 1mm, 1 degree) relative to all X-ray frame number.
As results of the experiments, 3D pose of the tibial component model for all X-ray images except for the first frame was full-automatically stably-estimated, and the automation rate was 80.1 %.
Conclusions
The proposed method by MAP estimation introduced the statistical motion model was successfully performed, and did not need labor-intensive manual operations for 3D pose estimation of tibial component.
For any figures or tables, please contact authors directly (see Info & Metrics tab above).
