Joint level-set shape modeling and appearance modeling for brain structure segmentation.

This paper presents a new fully automatic model-based segmentation algorithm, which combines level-set methods to model the shape of brain structures and their variation with active appearance modeling to generate images that are used to drive the segmentation. The new algorithm incorporates multi-modality images to improve the segmentation performance and the recursive least square (RLS) algorithm is adopted to minimize the difference between test image and the one synthesized from the shape and appearance modeling. When compared with manual segmentation, the 2D and 3D experiments demonstrate that the new algorithm is computationally efficient and robust and is promising for automatic segmentation of the lateral ventricles.