Study Of Liver Segmentation Algorithm Based On Active Contour Models

Segmentation of the region of interest in medical image is the key point of medical image analysis and understanding. Liver segmentation is the premise of liver cancer feature extraction and recognition. The accuracy of liver segmentation has important implications for liver cancer diagnosis and treatment and has recently become research priorities and hot spots in the field of biomedical engineering. But the conditions that there are many organs in abdominal cavity, liver is lack of good contrast with adjacent organs and it is difficult to find a clear boundary of liver, which bring a big difficulty to liver segmentation. With the development of the theory of curve evolution and level set method, image segmentation algorithm based on geometric active contour models can comprehensively utilize of regional and boundary information, therefore it is more suitable for the processing of medical image segmentation.The thesis will focus on how to improve active contour model segmentation algorithms, and applies them in liver segmentation, then extent to medical image segmentation. The thesis presents an improved active contour model method. The energy functional of the model contains both the driven term of global information but also the driven term of local information, so the segmentation result of our method is precise. In connection with the problem that level set method is sensitive to the initial contour, this thesis proposes a method by firstly using the region growing method to get the initial contour and then using improved active contour model method to segment the target area. The method cannot only get the precise segmentation results but also offset the weaknesses of these methods, for example, the region growing method is sensitive to the threshold and the level set method is time-consuming and sensitive to the initial contour. By considering the similar feature of CT image sequences adjacent image, this thesis presents a fast liver segmentation method based on the average shape of target area. This thesis which introduces the average shape of target area into the energy functional, can segment liver quickly and solve the problem that level set method is sensitive to initial contour. In addition, this method can detect the weak edge in the image and get an accurate segmentation result.In order to verity the effectiveness of our methods, this thesis makes many comparative experiments. The experimental results demonstrate that our methods can segment liver well while at the same time have a good robustness.