The advanced therapy equipment and surgical diagnose can be combined by computer aided surgery, which can provide safe, accurate and minimally invasive surgical therapy to the sufferers with utilizing the obtained information. The research content of computer aided surgery is related to digital image processing, computer graphics, computer vision, anthropotomy and other disciplinary field. In this dissertation, the key technologies of computer aided preoperative planning for the prosthetic replacement of femoral head and the plate fixation of diaplastic fracture, which are familiar prosthetic orthopaedics operations, are systematically studied. The research content mainly includes acquisition and preprocessing of the CT images, 3-D mesh segmentation of the key positions, sampling method and purification processing of the data point set, femoral head prosthetic modeling based on surface fitting and its model matching analysis, fractured bone splicing based on pre-registration and its plate prebending simulation.The preprocessing method for CT images of bones in the key postions is studied firstly. After extracting the region of interest (ROI) in the images, an improved Curvelet denoising method is proposed. In this method, the correlation of milti scale transform coefficients is effectively used to enhance the ability of Curvelet for extracting the edge detail information of the images, and the morphological method is utilized to restrain the scratch artifacts. The visual quality of the images is improved, which can promote the subsequent image analysis.Secondly, 3-D mesh segmentation of the key postions on the bone models is discussed. A method based on vertex normal feature is used to segment the facies articularis acetabularis and the fracture bone section. The median filtering in 2-D image is embedded to eliminate the isolated noise points, also the boundaries of the sub-meshs are smoothed. The segmentation effect and time efficiency of this method are all superior to the present typical algorithms.In the procedure of collecting data point set on the acetabulum, a special accurate segmention method is utilized to extract acetabulum binary images. A hybrid sampling method based on isometric raster sampling and spoke sampling is proposed to collecte the spatial data point set in the acetabulum image sequence. Afterwards, the elements in the data point set are extracted to eliminate the dirty data.The computer modeling of femoral head is the basis of computer aided repairing operation. In this dissertation, two prosthetic modeling methods based on sphere fitting and ellipsoid fitting are proposed. In the sphere fitting method, femoral head is considered as a regular sphere, and the refined data point set is used to solve the spherical center and the radius by fitting. In the ellipsoid fitting method, taking the anisotropy of ellipsoid into account, femoral head is considered as an irregular quadric surface, and the refined data point set is used to solve the location and angle parameters by fitting. Furthermore, the matching degree between the femoral head model and the acetabulum model is evaluated from geometry and stress. The experimental results prove that the suggested methods can achieve highly matching quality between these two models.Finally, the computer aided preoperative automatic splicing for fractured bone models in 3-D space is studied. Three coarse registration methods (respectively based on oriented bounding box, based on fractured bone axial line and based on fractured bone ridge line) are proposed. These methods can reduce the rotary and translational location difference between the two fractured bone models, and improve the efficiency of subsequent refined registeration. After analyzing the characteristics of traditional ICP algorithm, an automatic registration algorithm based on non-iterative estimation is proposed. Compared with ICP, the computational complexity and the time cost of this method are all lower. On this basis, the customized design and prebending simulation of the fixation plate for fractured bone models is implemented. After collecting the control points on the surface of models, a method based on immediately rendering the mesh surface of control points is utilized to construct the plate models. Therefore, various geometric parameters can be obtained before the surgical operations, and the operation time can be shortened.
|