Real-Time Dissectible Deformable Models Using High-Performance Breakable Shape Constraints 60

In this work we describe an approach to creating virtual dissectible models from color cryosection data, such as the images from the Visible Human Project, using high-performance deformable and destructible shape constraints. Prior research on deformable body fracture has largely focused on achieving physically accurate simulations, usually at the expense of real-time performance. We sacrifice engineering-level simulation accuracy, as would be expected in a surgical simulation, in favor of visually plausible animation. We use position based dynamics (PBD) to implement the dissectible model as a grid of constrained particles displayed as textured volumetric splats. Results are shown for a variety of real-world datasets, and we demonstrate fracture modes that are crucial to anatomy education applications: cutting, tearing and excision. A performance analysis shows that our compute shader implementation is capable of faster real-time performance than MPM and FEM, making it suitable for this application.