Liver is the largest solid organ of the body and plays a major role in metabolism. About one million people worldwide die from liver cancer each year and the disease ranks second in cancer mortality in China. The disease seriously jeopardizes health and life of people. Traditional liver surgery planning and evaluation only use case history and 2D medical image of patient, which is blind and unreliable when employed in complex liver surgery. Computer-aided liver surgery planning depends on medical imaging, image processing and computer graphics technology to help the surgeon with disease diagnostic and surveillance, oncologic resections, living-related liver transplants. It can reduce postoperative complications and improve the objectivity and survival rate of surgery. In order to explore anatomy of hepatic organ, tumor, vasculature and their spacial relationship, computer-aided liver surgery planning generally utilizes Multi-slice Spiral Computed Tomography (MSCT) data for analysis, processing and 3D visualization. The key problem of computer-aided liver surgery planning is medical image segmentation, which includes liver segmentation, liver tumor segmentation, hepatic vasculature segmentation and liver segments. Hepatic tissue segmentation in CT scans is the prerequisite for visualization, quantitative analysis and surgery planning. It has some difficult tasks including: (1) weak boundary between object and some adjacent tissues; (2) highly varying shape and appearance of object; (3) complex topology and high number of bifurcations of liver vessel; (4) image noise and artifacts. Efficient and accurate hepatic tissue segmentation becomes a challenging task that has attracted research attention recently. In this dissertation, we presented algorithms for segmentation of liver, liver tumor and liver vessel in CT scans and implemented the algorithms based on the medical imaging platform (MITK&3DMed) developed by our group. The main contents include: 1) We proposed a liver segmentation method using a statistical shape model (SSM) integrated with an optimal surface detection strategy. The approach firstly employs 3D generalized Hough transform to localize the SSM and finally employs a graph-based optimal surface detection algorithm to overcome the disadvantage of lacking flexibility of SSM. The method was tested on MICCAI 2007 liver segmentation challenge datasets. Experimental results show that our method can explore liver shape prior abundantly and adap...
修改评论