Takaharu Yamazaki PhD, Presenter: Nothing to Disclose

Tetsuya Tomita MD, Abstract Co-Author: Nothing to Disclose

Tetsu Watanabe MD, Abstract Co-Author: Nothing to Disclose

Hideki Yoshikawa MD, Abstract Co-Author: Nothing to Disclose

Kazuomi Sugamoto MD, Abstract Co-Author: Nothing to Disclose

The purpose of this study is to accurately determine the three-dimensional (3D) normal knee kinematics, under dynamic flat-panel detector (FPD) system which can overcome the problem of bone edges detection, using a robust feature-based 2D/3D image registration technique.

First, the 3D bone models of femur and tibia/fibula were created from CT scans of the subject’s knee. CT scans used a 512x512 image matrix, 0.63 mm slices. Next, dynamic FPD images during the subject’s knee motion were acquired as serial digital images (750x750x12 bits/pixels, 5.0 frames/sec). To extract the knee contours from the acquired image, a Canny’s edge detector was applied, and as a result, use of dynamic FPD images enabled us to easily detect bone edges. Finally, in order to determine 3D pose of the subject’s knee model from a FPD image, a robust feature-based 2D/3D image registration technique which use 2D knee contours images and 3D bone models was developed. In order to validate the accuracy of 3D pose estimation of the knee bone using the developed technique, phantom experiments were performed using knee images taken in typical 5 different poses. The accuracy was assessed by comparing the estimated relative pose between femur and tibia/fibula with the relative pose determined by 3D model from CT images. In the clinical applications, we performed 3D kinematic analysis during knee squatting motion for 6 healthy male subjects.

The results of phantom experiments showed the relative poses errors of femur with respect to tibia/fibura were 0.59 mm and 0.99 deg except for medial-lateral translation error. In the clinical results of squatting motion, the femur exhibited average 21.0 deg of external rotation relative to tibia during knee flexion (average from -2.8 deg of full extension to 146.6 deg of full flexion). Then, the contact position of both condyles sifted posteriorly average 13.4 mm medially and 24.6 mm laterally, respectively.

The 3D determination of normal knee kinematics was successfully performed using the dynamic FPD system and a robust feature-based 2D/3D image registration technique. This technique could be useful for analyzing 3D knee kinematics and for diagnosing knee joint.

In vivo quantitative assessment of 3D dynamic knee kinematics is important for understanding the effects of joint diseases and dysfunction, and is recommended in diagnosis and surgical planning.

Yamazaki, T, Tomita, T, Watanabe, T, Yoshikawa, H, Sugamoto, K, Three-dimensional Determination of Normal Knee Kinematics Using Dynamic Flat-Panel Detector System.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12036258.html