A coupled implicit shape-based deformable model for segmentation of MR images

In this paper, a new coupled implicit shape-based segmentation algorithm is proposed for medical image segmentation. In the method, both region-based and statistical model-based curve evolution algorithms are jointly used to match the object in a new input image. Compared to the previous method that solely uses statistical shape models, our new algorithm is able to match the boundaries of the object shapes more accurately and at the same time, it maintains similar robustness since the same shape prior information is used to regularize the object shapes. Experiments on segmenting the ventricle frontal horn and putamen shapes in MR brain images confirm that the proposed algorithm yields more accurate segmentation results.