Improvements of GPU Implementation of Nonlinear Soft Tissue Deformation with CHAI 3D

Abstract

The contradiction between the speed and the accuracy of soft tissue deformation simulation has challenged developers for decades, especially for nonlinear deformation. In this paper, we propose improvements of the GPU implementation of nonlinear soft tissue deformation with CHAI 3D by decreasing the calculation scale and reducing data transfer between device and host. Soft tissue is divided into deformation area and non-deformation area, and points and tetrahedrons in the deformation area are calculated during the deformation. In addition, we redesign the collision detection of CHAI 3D and implement it with CUDA in parallel. By using VBO, OpenGL displays the soft tissue based on data resided in graphics memory directly. As a preparatory work of FEM, we also present a convenient method to divide the soft tissue into tetrahedrons by TetGen.