Study On Key Technologies Of Medical Image Segmentation And Virtual Surgery

Medical Imaging has played an important role in modern medical treatment and diagnosis, many surgical operations and disease diagnosis procedure need the participant of medical imaging technology. The application of medical imaging technology in modern medical treatment has no more limited in common operations like reading the picture and reconstructing the 3D image. In order to dig out deep-seated infomation and provide accurate and reliable foundation for diagnose and curation, people begin to analysis and understand the obtained images more comprehensively by incoorporating many advance technology. The medical image process and analysis technology not only involves many basic contents like the image segmentation, image feature extraction, 3D reconstruction etc, but also include some practical contents like the computer aided diagnose, virtual surgery, surgery guidance etc. This dissertation selects the cortex segmentation, objects motion in 3d reconstruction and presurgery design system as the key research contents, and takes the construction of practical presurgery design system as the ultimate goal.The cortex has very complex structures, each part of them are correponding to part of human body functions. The cortex segmentation and representation are the important phase to analysis the cortex and unveil the function mappings, meanwhile the cortex segmentation and representation are also the nut to the medical imaging field. In this dissertation, a adaptive ribbon model based on levelsets is proposed to tackle the cortex segmentation and representation problem. This model is naturely a hybrid adaptive levelset model combining both the merits of region and boundary segmentation. The regionbased levelsets are used to represent the region information, and region information act as levelset evolving motivition term. The statistical coupled edges detector is used to describe the boundary information, and the boudary information acts as a stopping term to the evolving process. This model is provided with both the robustness of region model and the preciseness of boundary model. According to the fact that the cortex has the similiar thickness, the model takes this fact as a prior knowledge and shapes into adaptive ribbon model. Besides, the bias field description is added to overcome the influence of the possible bias field in MR image. To demonstrate the proposed model, the simulated brain images and real brain images are used for testing, and comparation to other methods are made. the experiment results show that adaptive ribbon model is more robust and accurate than other models.Image intensity overlapping between brain tumor and other normal tissues make the brain tumor segmentation a tough thing. For the sake of overcoming this brain tumor segmentation difficulty, a framework based on viscous fluid model registration for brain tumor segmentation is proposed.This framework consists of registration building on viscous fluid...