Hybrid framework for medical image segmentation

Abstract

Medical image segmentation is essential step for many image processing applications. In this paper, we present a hybrid framework designed for automated segmentation of radiological image, to get the organ or interested area from the image. This approach integrates region-based method and boundary-based method. Such integration reduces the drawbacks of both methods and enlarges the advantages of them. Firstly, we use fuzzy connectedness method to get an initial segmentation result and homogeneity classifier. Then we use Voronoi Diagram-based to refine the last step's result. Finally we use level set method to handle some vague or missed boundary, and get smooth and accurate segmentation. This hybrid approach is automated, since the whole segmentation procedure doesn't need much manual intervention, except the initial seed position selection for fuzzy connectedness segmentation. © Springer-Verlag Berlin Heidelberg 2005.