Direct Volume Rendering Methods for Needle Insertion Simulation

This thesis presents a simulation framework and newly developed methods for visuo-haptic rendering of a virtual liver puncture intervention. Visuo-haptics include both visual representation and haptic rendering using a haptic input device for force display. The framework includes methods and algorithms to simulate a medical intervention that consists of palpation, ultrasound probing and X-ray imaging. Using these components, a percutaneous transhepatic cholangiodrainage can be performed in virtual reality with the aim of providing an environment for training and planning. The central aspect distinguishing the innovative approaches from comparable state of the art methodology is the focus on methods that use direct volume rendering of medical computed tomography image data. Therefore, no creation of an intermediate surface representation of organ tissues has to be performed. These surface representations are needed in comparable approaches in order to perform visualization, force feedback computation and soft tissue simulation. To create these surface representations, it is necessary to perform a time-consuming segmentation process. This process took more than 60 hours in the framework’s predecessor. It is essential to reduce this time in case a patient specific scenario should be prepared based on new image data. In the presented framework, the segmentation is reduced to structures that are central to the intervention. The resulting partial segmentations and patient image data are then rendered visually and haptically during run-time by adapted and newly developed rendering approaches. Visual direct volume rendering is realized by ray casting of the volume data. A soft tissue simulation component is included into the framework by computation of local deformations on the regular grid of the volume data. The resulting deformed image is then considered in the volume rendering process. Furthermore, respiratory motion can be visualized by the framework and is integrated into the haptic algorithms. This is done by using 4D CT image data and creation and application of resulting motion models. For achieving real-time rendering capability, the visualization methods are implemented using Nvidia CUDA. First of all, this thesis summarizes the newly developed methods and obtained results and then analyzes the components of the framework. It is shown that the methods and their implementation fulfill the requirements with regard to the reduction of segmentation effort, real-time capabilities and plausibility of visuo-haptic rendering.