RESUMEN
BACKGROUND:Due to anisotropic CT volume data, triangular meshes extracted from bone CT images often contain staircase surface, which wil affect the subsequent medical diagnosis. OBJECTIVE:To reconstruct patient-specific bone model based on digital geometry processing technique. METHODS:Firstly, registration was performed in image registration algorithm based on mutual information for bone CT slices, and then contours were extracted by image segmentation and a stack of contours were converted into point clouds. The normals of point clouds were estimated based on Gaussian weighted principal component analysis and the noise from point clouds was removed by trilateral filtering. Final y, bone triangular meshes were constructed by adaptive spherical cover. RESULTS AND CONCLUSION:In this paper, the proposed method could generate smoothing bone surface meshes, triangular mesh shape which was formed by the rules and adaptive distribution, for finite element analysis, computer aided manufacturing and three-dimensional printing to provide accurate three-dimensional models.