Adaptive soft tissue deformation for a virtual reality surgical trainer.

Real time tissue deformation is an important aspect of interactive virtual reality (VR) environments such as medical trainers. Most approaches in deformable modelling use a fixed space discretization. A surgical trainer requires high plausibility of the deformations especially in the area close to the instrument. As the area of intervention is not known a priori, adaptive techniques have to be applied. We present an approach for real time deformation of soft tissue based on a regular FEM mesh of cube elements as opposed to a mesh of tetrahedral elements used by the majority of soft tissue simulators. A regular mesh structure simplifies the local refinement operation as the elements topology and stiffness are known implicitly. We propose an octree-based adaptive multiresolution extension of our basic approach. The volumetric representation of the deformed object is created automatically from medical images or by voxelization of a surface model. The resolution of the volumetric representation is independent of the surface geometry resolution. The surface is deformed according to the simulation performed on the underlying volumetric mesh.