The selection of a patient-specific optimal implant with respect to shape, size, and position during preoperative planning requires the consideration of geometrical and biomechanical criteria. In particular, the selected implant should lead to a physiological stress distribution in the bone, in order to avoid a degeneration of bone tissue and eventually a loosening of the implant or a fracturing of the bone.

In this talk, I present the first planning system that allows for an implant selection based on the patient-specific prediction of the stress distribution. The system provides a virtual 3D planning environment, where the physician can interactively insert implants into the bone. During this process, the system instantaneously computes and visualizes the resulting stress distribution. In my talk, I first address the interactive simulation of the stress distribution by means of a highly efficient hexahedral multigrid finite element method, including the implementation of this method on the GPU using the CUDA computing API. I then discuss the interactive visualization of the resulting 3D stress tensor fields, including the comparative visualization of these fields to allow for a clear rating of an implant with respect to stress shielding.

Finally, I present methods for the interactive visualization of the spatial distances between bone and implant that support the validation of the specific geometrical criteria. All simulation and visualization methods have been implemented on the GPU to exploit its massive computing power and memory bandwidth.