Research of New Medical Volume Visualization Methods: Application in the Kidney Preoperative Planning System
Hui Nee Tang · HAL (Le Centre pour la Communication Scientifique Directe) · 2008
This dissertation focuses on the main elements of a scientific visualization tool and takes a kidney preoperative information review system as a special application example to introduce the corresponding algorithms. Our research work followed the essential stages of the design of the kidney visualization system: registration, segmentation and visualization. The CT uroscan consists of three to four time spaced 3D acquisitions, which give complementary information about the kidney anatomy. In order to bring these acquisitions into spatial alignment, a kidney centered registration method which is realized by local mutual information maximization is proposed. In order to illustrate the information contained in the spatial aligned volume, an acquisition level intermixing method is proposed, which intermix the several component volumes at the earliest stage. The first step for the acquisition level intermixing is a vectorial volume classification. We proposed a neighborhood weighted Gaussian mixture model, which involves the spatial information into the classification process. Then, several possible rendering techniques that can be adapted to this situation are presented and compared. For surface based volume visualization methods, mesh simplification is a usual way to improve rendering speed. The simplification metric is a key issue of a simplification algorithm. Two new mesh simplification metrics are proposed. They are based on surface moments and volume moments respectively. Although these algorithms are introduced in the framework of the kidney visualization system, they are not limited to this system and can also be adapted to other applications.