Liver segmentation with shape-intensity prior level set combining probabilistic atlas and probability map constrains

In this paper, we develop a 3-D segmentation framework for fully automatic liver contrast-enhanced CT images that uses shape-intensity prior level set combining probabilistic atlas and probability map constrains. We first weight all of the atlases in the selected training datasets by calculating the similarities between the atlases and the test dataset to dynamically generate a subject-specific probabilistic atlas for the test dataset. Based on the generated probabilistic atlas, the most likely liver region (MLLR) of the test dataset is determined. Then, a rough segmentation is performed by a MAP classification of probability map. The final result is obtained by applying a shape-intensity prior level set segmentation inside the MLLR implemented by narrowband technique. We validate our method on 10 liver cases by comparing our segmentation result with manually traced segmentation result. Experimental results show the effectiveness of the proposed method.