|
Virtual surgery is an emerging field which inosculates modern medicine and virtual reality technology, and it consists of computer graphics, modern medicine, artificial intelligence, image processing, Biomechanicsand so on. A complete virtual surgery system contains model construction, deformation simulation, cutting, collision detection, graph display and force feedback. Thus, it is a huge challenge for developers to construct such a large system completely. In order to accelerate the research of virtual surgery system, international research organizations and scholars have developed lots of development packages and libraries that contain basic algorithms and modules used in virtual surgery, which make it possible for developers to concentrate on a field and make an intensive study of the field. The representatives of such packages and libraries are SOFA, CHAI3D and SPRING.By using explicit finite element method, the paper researches the nonlinear soft tissue deformation with haptic feedback within CHAI3D. The deformation process can be divided into lots of discrete time points, and at each time point, the status of soft tissue is described with total Lagrange method, that is,it refers to un-deformed status. Using the difference method, the constitutive equation can be described as an explicit equation of displacements, and the previous and current displacements are the only variables saved during the deformation, which reduces the amount of storage and makes it possible to solve the equation in parallel.The whole deformation process consists of initialization phase and cyclicallycalculation phase. In initialization phase, we read the deformation model, compute the variables which do not change during deformation and save them in GPU, allocate memories in GPU and execute other preparations. In cyclically calculation phase, the collision detection gets the triangle containing collision point when the virtual device interacts writh soft tissue. According to collision triangle, we collect the deformation area, and compute displacements and reaction force on all the points and tetrahedrons in the deformation area. At last, we feedback these displacements and reaction force to user by visual display and haptic device.For model structure, we employ3D modeling tools, such as3D MAX, Maya, to construct3D model based on medical data, then divide them into tetrahedrons by TetGen, and save them and other information used for deformation as XML files.Soft tissue deformation has faced with the contradiction between simulation accuracy and solution speed. Thus, this paper uses GPU to deal with the time-consuming calculations, for the computing power of GPU is several dozen times, even several thousand times than that of CPU. As most of the calculations are computed by GPU, and the results of computations are saved in graphics memory. Thus, for reducing the transmission of data between the host and GPU, we display the data in graphics memory by the interoperation between CUDA and OpenGL, in addition, we designand implement the collision detection of CHAI3D in parallel, thus, it can use the data in graphics memory to detect a collision, which improves computational efficiency. In addition, the soft tissue is divided as deformation area and un-deformation area, and the calculation is only executed in deformation area, which reduces the calculation scale.
|