Consistent Mesh Generation for Non-binary Medical Datasets

This paper presents a new approach for constructing a consistent (non-manifold) mesh based on non-binary datasets. Mesh generation is the method of choice if a surface is extracted based on labeled volumetric datasets. In most cases, the traditional Marching Cubes algorithm or one of its extensions is applied for generating meshes of binary labeled input data. However, for different — especially medical — applications, non-binary datasets with more than two differently labeled voxels can be adjacent in a rectilinear grid. For these cases, the traditional Marching Cubes algorithm fails. Additional to the mesh generation algorithm, an extension for a well-known geometric filter is presented avoiding ridges in the mesh.