Three-dimensional (3D) surface construction from computed tomography (CT) slices has received extensive attention as many biomechanical analyses and designs rely on the development of a reliable 3D model. This study proposes a method for constructing the triangular models of anatomic structures from two-dimensional slices of CT images. A region growing algorithm is developed for the segmentation of the volume data. Each voxel on the volume data represents a pixel of 3D images. A modified marching cubes algorithm, a 3D reconstruction algorithm, is then employed to establish the triangular model. Problems caused by imprecise image data are mitigated. In addition, a data reduction algorithm that combines a pair of voxels along each coordinate direction is developed, in which piecewise linear interpolation is implemented to maintain the accuracy of the reduced model. Several examples are presented to demonstrate the feasibility of the proposed method.