Research On The Acquisition Of Individual Informations Of Livers And Model-Driven Methods In Virtual Surgery

The personalized modeling and simulation of interactive human liver models is one of the key technologies of the liver virtual surgery, it can give a clear intuitional and stereoscopic view of the anatomical structure of a liver, provide evidences for the treatment of hepatic disease and the establishment of the operation strategy, and is of important significance to the study of the anatomical structure of a liver and the liver virtual surgery. The majority of existing liver modeling methods are based on the 3D reconstruction of CT images, which will lead to the problems that we have to achieve the reconstruction for every CT series and the models acquired through the reconstruction of CT images are of low accuracy because of the low resolution of CT images.Aiming at the problems described above, we focus on the research of the modeling of the surface of 3D liver models, and present an individual liver modeling method based on data driving. Firstly, we establish a 3D liver model with a higher degree of accuracy based on the VHP liver sections, then we obtain the individual informations from the CT sections of different individuals, and drive the liver model to deform to get the final liver model with a higher degree of accuracy as well as the individual informations of the specified individual. This paper’s research work mainly includes:1. Establish a standard 3D liver surface model with a higher degree of accuracy. Since the VHP sections are of higher degree of accuracy than CT sections, we use the segmented VHP liver sections to achieve the construction of a standard 3D liver surface model with a higher degree of accuracy through 3D reconstruction methods. This is an important preparation for the study of our paper, and the model acquired can be reused in the process of subsequent personalizing.2. The acquisition of individual informations of specified livers. Based on the problems of existing segmentation methods of liver CT images, we combine classical region filling algorithms with the B-Snake model algorithm, classifying the control points of B-Snake model using the intensity distribution property inside and outside the liver regions, and making use of the correlation of the adjacent sections to recognize and segment the multi-regions of a liver in a section.3. Personalized data driven method to the 3D liver model.Before the model driven deformation of the liver model, we should firstly establish the dense correspondence between the CT contour point sets and the liver model. In order to reserve the existing tomographic structure on liver model surface, we present an improved ICP algorithm based on closest contour line, finding the closest contour lines to get the matching points and limiting the deformation to the horizontal direction only.After the dense correspondence between two point sets is calculated, in order to improve the computing accuracy, the surface of the liver model is divided into several regions and every region corresponds to a RBF system, then the model is deformed according to the RBF system and the constraints of the CT point sets, finally, the common boundary of every two adjacent deformation regions is smoothed throuth smoothing methods of 3D meshes.Finally, we designed and developed a system for the personalized modeling of digital liver models, in which we integrate the algorithms described above and demonstrate the feasibility of the algorithms. The experimental results show that the modeling method we present is of high efficiency and the liver model we finally get with the algorithms described above is with high precision, a smooth surface and personalized informations of specified individuals, and it can offer a reliable model foundation for the subsequent studies of 3D solid texture mapping as well as the construction of blood vessel tubes, and have bright prospects in future application.