Landmark based liver segmentation using local shape and local intensity models

A 3D medical image segmentation algorithm is presented which models an object as a set of landmarks augmented with local appearance models similar as with Active Shape Model segmentation methods, but instead of applying a global shape model, multiple local shape models are used. A global cost function incorporating local in- tensity and local shape is optimized iteratively. A first implementation of the method is validated for the segmentation of the liver in contrast enhanced CT images and demonstrates that the method has potential. This work participates in the MICCAI 2007 Grand Challenge on 3D segmentation (1).