Towards a GPU-based Accelerated Surgery Simulation System

Rui Hu, Kenneth Barner, and Karl Steiner


Surgery simulation, Graphic processing unit, Compute Unified Device Architecture, Point based animation


Modern surgery simulators offer surgical residents and trainees an immersive environment to acquire then evaluate their skills. In many cases, a simulator includes several key components: soft tissue deformation, collision detection, contact force distribution and haptic force feedback. While a highly accurate, interactive simulation environment requires intense calculation resources, the traditional CPU may not provide enough computational power for such a resource-demanding application. As an alternative, this paper explores the adoption of the graphic processing unit (GPU) as the main processing tool to address the computational bottleneck existent in the CPU. We have set up a surgery simulation system using point-based animation as the main deformation algorithm. An image-based collision detection and response algorithm has been improved and implemented to provide accurate contact force. A volume-based haptic feedback algorithm is also applied to the simulator. Calculation for these three steps is carried out on nVidia's Compute Unified Device Architecture (CUDA). Compared with the traditional CPU approach, the processing time for each frame on our implemented system has been significantly reduced. For the deformation, the frame rate has increased 10 times on the GPU. This GPU-based simulation framework will allow additional models with high complexity to be included in the simulation. This will not only improve the realism but also the accuracy of future surgery simulation procedures.

Important Links:

Go Back