Alternatives for the generation of triangle meshes to represent bone fragments

Abstract

The use of computational techniques to help specialists in the treatment of bone injuries and diseases is widespread. These techniques usually require 3D triangle meshes representing bone fragments as input. On the one hand, bone fragments can be visualised in a more effective way. On the other hand, computer-assisted techniques are provided with models that make the application of geometric algorithms easier. The suitability of the method utilised to generate the triangle mesh will depend on the computational techniques in which it will be used. In this paper, a study is performed in order to test the applicability of some mesh generation methods to model bone fragments: Marching Cubes, Ball-pivoting, Poisson reconstruction and Algebraic Point Set Surface. In a first step, bone fragments were segmented from CT scans, and external contours were extracted from the segmented regions. After that, the vertices of each external contour were grouped in order to form the point cloud used as input by the mesh generation methods. Finally, these methods were tested and their main benefits and shortcomings were exposed. The obtained results manifested the performance of each method depending on the fracture type and the special features of the data obtained from CT stacks. In addition, the suitability of each method to be used in computer-assisted medical procedures was analysed and discussed.