Ct Tumor Sequence Image Segmentation

Medical image processing technology is a trend of research combining computer graphics technology and medical knowledge. It is usually the basic of the subsequent operation, however, the medical image segmentation technologies have not formed a unified and effective segmentation method until now, which leading to the emergence of a new task, and a new segmentation method is also needed to re-development. These new segmentation methods might be some new methods different from previous segmentation methods or might just the improving and deformation of previous segmentation method, but they have a huge contribution to make new solutions to the problem as well as achieve advances in technology.As computed tomography (CT) technology is more and more popular and mature, it has become the doctors’main means of observation for tumor growth. Timing regular observation can accurately grasp the tumor size and status, providing data for the efficacy of the drug and the effect of chemotherapy. However, a large number of CT images thus emerged. The manual segmentation consumes enormous time and the existence computer image segmentation methods would not be suitable because of not taking these new features of CT sequence images into account. For this new task, we developed a new segmentation method. Its main idea is registration before segmentation. We use registration methods to eliminate the displacement and rotation caused by machine parameters and segmentation method to fix the deformation caused by the tumor itself. In the registration process, we combine the outer contour registration with the dual template matching and fast deformable registration (FFD).After the registration process, good initial contour is ready for segmentation, in order to make full use of the initial contour in the segmentation stage, we use the improved Snake algorithm and improve marker watershed method. We list part of the experiment data and introduce a segmentation evaluation method. In the end of the article, we compare the algorithms we develop with the previous algorithm.