| Summary: | 碩士 === 國立臺灣大學 === 醫學工程學研究所 === 91 === An accurate, non-invasive method for in vivo bone position determination is essential for the study of joint kinematics. Previous studies using skin markers on body segments to determine the underlying bone positions are prone to errors due to undesired motion between the skin and bone. Attaching external markers directly to the bone by using bone-pins provide accurate measurements, but it is invasive and risks infection. Traditional x-ray imaging allows non-invasive and accurate measurement but provides only two-dimensional (2D) static information. Fluoroscopy offers a means of measuring dynamic bone movement free of skin movement artefacts, but is limited to planar information as well. Several model-based methods that combine 2D fluoroscopic images and three-dimensional (3D) computer model of the total knee replacements (TKR) have been shown to be feasible and reasonably accurate in measuring TKR kinematics. However, these methods were specifically for replaced joints. A method for the kinematic measurement of joints without implants has yet to be developed. The purpose of the study was to bridge the gap by developing a model-based method for the determination of 3D dynamic skeletal motions in vivo. The proposed model-based method is based on image registration techniques and determines bone position by using 3D computer model and measured 2D fluoroscopic images of the bone.
The video-fluoroscopy system needs to be calibrated before subsequent analysis, and image distortion coefficients and X ray source position are calculated. A five order polynomial function is used for correcting images, and the system internal parameters are estimated by optimization. The 3D computer knee geometric model is established by serial CT transverse plane images. The knee motion images are recorded by fluoroscopy and the contours of the bone are segmented and corrected distortion. The 3D position and orientation of the bone are calculated by optimization of minimizing errors between the projected model contour and segmented bone contour.
Computer simulation and experimental validation are required for determination of the accuracy of present study. TKR and patella bone are acquired for experiments and relative accuracy is determined. The accuracy is smaller than 1.5 degrees and 2.3mm of TKR at orientation and translation respectively, and 2degrees and 1.7mm of the bone. In the application of clinical studies, the results are not good as well because of the unclear of the bone contours from fluoroscopy. The study develops a new method combined with computer bone model and fluoroscopy for knee kinematic measurement, and is helpful for the study of joint kinematics in vivo.
|
|---|
