Conformal multi-material mesh generation from labelled medical volumes

Generating accurate volume meshes from segmented image volumes is important in certain medical applications. One application is finite element analysis (FEA) for orthopaedic pre-operative planning. The resulting FEA models can be used in patient-specific, bio-mechanical simulations and implants positioning. Currently available software packages for volume meshing of segmented data can generate accurate, usable meshes. However many of them cannot create feature-adaptive volume meshes of multiple intersecting structures while maintaining high precision. An approach that respects both these requirements is based on a dynamic particle system. In this paper, we propose a system that specifically includes the ability to handle multiple labels and their interfaces. Dynamic particle systems are computationally expensive and may be slow. As an improvement to our reference method, we implement a fast Integer Medial Axis algorithm. Furthermore, we propose an improved local triangulation scheme that may enhance the ability to model sharp features with a given particle count. Strict theoretical sampling restrictions limit the flexibility of Delaunay-based meshing methods. We suggest that using a local reconstruction scheme is a good approach to decouple the system from strict sampling limitations. We evaluate our system by comparing it to a reference particle-based meshing implementation. Results show that the choice of the Medial Axis Transform (MAT) methods greatly influences quality and speed of featureadaptive volume meshing.