Echtzeiffähige Gewebemodellierung für chirurgische VR-Trainingssimulatoren

Surgical interventions on humans need to be performed with a high grade of efficiency, security and care. A major problem is the absence of possibilities for learning and training of surgical operations under realistic conditions. Virtual Reality (VR) based training systems seem to be a good alternative to classical training methods, which allow unlimited repeatable training with objective assessment. VR based training systems mostly offer a realistic environment with haptic input devices and virtual surgical interactions with deformable or static organ models. The level of realism of a VR based surgical training system depends mostly on the quality of the tissue model. Tissue modelling covers the topics of geometrical model creation using CT and MR data, the physical simulation of object deformation, the realistic visualisation and the surgical model interactions. In this project, we developed new methods for modelling, visualisation and the simulation of tissue deformation. As a development environment we used "VSOne", a VR based training system for minimally invasive surgery developed at Forschungszentrum Karlsruhe. We developed a software module for modelling deformable objects with implicit surfaces, which are most suitable for inexperienced modellers without any geometrical knowledge. This module supports also the geometrical modification of existing 3d models. Special model 10-procedures enable to convert static models into a deformable model representation and to import them into the simulation kernel of "VSOne". The support of deformable models with arbitrary triangle meshes within our simulation software allows the model exchange with other developers and helps to build up a world-wide network of model databases. Further development was done in the area of realistic visualisation of virtual tissue for real-time applications. A real-time texture composition algorithm enables to map basic, structure and lighting textures onto the 3d objects utilizing modern graphics card's multi-texturing option, but also supports older hardware with single texture-units. Real-time simulation of soft tissue deformation is mostly done using mass-spring models. As an alternative we use the Fast-Finite-Element-Method (FFE). We extended the existing implementation with several new features like object-oriented and adaptive calculation of the stiffness matrix and the possibility for hybrid deformation simulation. A two-step FFE-simulation was examined for suitability in a VR based surgical training system.