Segmentation of 3D CT volume images using a single 2D atlas

Abstract

Segmentation of medical images is an important first step in the analysis of medical images. A lot of research has been performed on the segmentation of complex CT/MR images using the atlas-based approach. Most existing methods use 3D atlases which are more complex and difficult to control than 2D atlases. They have been applied mostly for the segmentation of brain images. This paper presents a method that can segment multiple slices of an abdominal CT volume using a single 2D atlas. Segmentation of human body images is considerably more difficult and challenging than brain image segmentation. Test results show that our method can handle large variations in shape and intensity between the atlas and the target CT images. © Springer-Verlag Berlin Heidelberg 2005.