Research On The Real-time Large Scale Deformation Method In Virtual Knee Joint Surgery

Virtual surgery is an application of Virtual Reality (VR) in medicine. VR-based surgical simulator has shown good prospect in application. Visual and haptic feedbacks are two important ways of interaction between visual surgery system and its user. Realistic and real-time visual and haptic feedbacks may enhance the immersion of a visual surgery system.Deformation of human soft tissue is a common phenomenon in real surgery. Generally, there are two types of deformation in surgery. The one is small deformation caused by contact between medical instruments and human organs, such as sunk or bulging of human skin. The other one is large deformation induced by body moving like articulated deformation. Simulation to these deformation phenomena is essential to the visual realism of a simulator. Simulation of deformation in virtual surgery must be both realistic and real-time.We in this thesis research the large deformation used in virtual surgery. Different from small one, large deformation performs large scale deformations and hence involve the entire mesh model. Physically-based models, especially the linear elastic models, have been widely employed in simulating local and small deformation on surface meshes due to their similarity. However, when used to simulate large deformations, nonlinear elastic models should be considered since linear models often present many artifacts. Nonlinear models are often too expensive to ensure real-time simulation. Moreover, the problem scale of large deformation is often quite large. All these facts prevent us from using physically-based models.We find geometrically-based methods are suitable for real-time large deformation modeling due to their efficiency, robustness, and physically plausible feature, which can meet the requirements of virtual surgery. We use the geometrically-based methods to simulate large deformation in virtual surgery. Geometrically-based deformation methods have been extensively researched in the context of computer animation.We propose a skeleton-driven deformation method based on smooth interpolation of transformation with cylindrical coordinates. We first extract the skeleton of a mesh. The deformation of the skin mesh is driven by the moving skeleton. The surface details can be preserved well due to our smooth interpolation scheme. Besides, our model is very efficient and can be performed in real-time.Gradient domain method based on differential coordinates is another popular geometrically-based mesh deformation method. We augment the existed gradient domain method with two features and propose a improved gradient domain method. To make the deformation material-aware, we adjust the weights in the energy function. We add a virtual node near the model part that may introduce the most volume degrading during deformation, such as joint part in the articulated deformation. By constructing a volume graph around this virtual node, our deformation is calculated using the volume differential coordinates, which can effectively preserve the volume of the joint part of a mesh.We introduce the application of the proposed methods in our virtual surgery system. The results show the effectiveness and efficiency of our deformation method.