Research Of Virtual Endoscopy System Based On VTK

Virtual Endoscopy incorporates with virtual reality, computer graphics, medical image processing and scientific computing visualization techniques, which is an application of virtual reality in modern medicine simulating optics endoscopy by use of medical images. With the advent of high resolution medical images (such as CT, MRI etc.) and the progress of virtual reality technique, virtual endoscopy is developing rapidly. It has been one of the most flourished fields during the application of virtual reality to practice. Virtual endoscopy is a complete non-invasive, untouched medical diagnosing method avoiding the risk of invading body while the actual optical endoscopy is used. It is becoming more and more widely used in many medical fields. The research objective of virtual endoscopy system is to provide trusty means for medical diagnoses, and to spread the new application fields into assistant diagnosing, surgery planning and training.Several key techniques existing in virtual endoscopy system are studied in this thesis. Firstly, the medical image pre-processing and tissue segmentation techniques are surveyed. In the aspect of medical image pre-processing, the anisotropic diffusion filtering algorithm is realized and parallel anisotropic diffusion filtering algorithm based on CUDA is proposed which the noise is smoothed while the characters of image boundaries and details are preserved. In the aspect of tissue segmentation, we realize the region grow algorithm, accurately extract the body colon data which is used in the test of centerline extraction, and propose the C-V level set segmentation based on CUDA which solves the shortcoming of computing consuming in level set method, and can greatly reduce the computing time while preserving the qualitative results. Secondly, in order to realize the path-planning in virtual endoscopy system, surveys are made on definition and features of center path as well as its extracting technology. A centerline extraction algorithm based on the distance from boundary field is realized and improved, which used distance transform technique. The propagation of the maximum spanning tree is accelerated using the multi-resolution and the priority table so as to greatly reduce the time of centerline extraction. Thirdly, in the study of scene rendering in virtual endoscopy system, an introduction is made on surface rendering and volume rendering algorithms in 3D regular data set, and 3D medical image visualization is realized by use of visualization toolkit VTK. Finally, based on the deep study of the above key techniques, a virtual endoscopy system with interactive functions is developed, the user interface is provided, the correctness and effectiveness of the relevant algorithms are demonstrated and prospects are made on several aspects in further study.