Design And Implementation Of Liver CT Images Interactive Segmentation Based On 3D Snake Models

Segmentation of liver CT images plays an important role for the liver disease diagnosis and treatment in clinical medicine. Fully automatic segmentation algorithms existing now are prone to arise segmentation errors for liver images, including false positive errors and false negative errors. In response to this problem, an interactive segmentation solution for liver CT images based on 3D Snake models is proposed to make effective correction for fully automatic segmentation algorithms and segmentation errors.Liver in the abdominal CT images is more complex, three problems should be solved when applying 3D Snake models to CT images to interactively segment liver. First, how to define an appropriate initial surface contour, through which segmenting liver under the evolution of the 3D Snake; Second, how to add external force to guarantee the best correction; Third, how to avoid that surface contour in position without adding external force where the original segmentation is right is pulled to the impurities where local edge is more strong after adding external force. For the first problem, the active contour models based on global optimization algorithm for pre-segmentation of liver initially is chosen to get a result mask data of pre-segmentation, on which we make surface sampling. Then Delaunay triangulation technique is used to establish liver surface contour model to get initial surface contour of liver. For the second problem, an scheme of interactively adding forces is proposed. For the third problem, we solve it from two points: points on the surface contour and image force. For points, we divide it into two parts: evolution points and fixed points. For image force, we strengthen image force of liver edge on the basis of enhancement liver data through data preprocessing. Until getting the final satisfaction segmentation of liver under the evolution of the 3D Snake.Experiments show that the proposed algorithm of interactive segmentation can get a good correct effect on error result of pre-segmentation. Especially, the effect will be more obvious for the cases where degree of error segmentation is large relatively, liver edge is clear, edge varies small, and impurities around is less.