Exploring a graphic cluster based solution for real-time virtual surgery

Surgical simulators have been continuously gaining popularity in the medical world. As simulated morphologies become increasingly complex, there is a growing demand for simulators to do fast and precise calculations in order to allow the user to interact with the system in real-time. Due to performance reasons, many existent surgical simulators are based on simple, fast mass-spring deformable models. However, even physically based models such as these, display slow convergence behavior as model resolution and size increases. Parallelization can be used in order to improve the performance of surgical simulators and to allow improved realism of tissue deformation. This paper explores a graphic cluster based solution to accelerate calculations for mass-spring models. Acceleration is obtained by harnessing the parallel processing power of the GPU and by dividing computational effort among more computers, organized in cluster architecture. The solution uses the new OpenCL standard for General Purpose GPU programming, OpenMPI for distributed computing and VirtualGL for remote visualization and interaction.