Segmentation of Left Ventricle via Level Set Method Based on Enriched Speed Term

Level set methods have been widely employed in medical image segmentation, and the construction of speed function is vital to segmentation results. In this paper, two ideas for enriching the speed function in level set methods are introduced, based on the problem of segmenting left ventricle from tagged MR image. Firstly, a relaxation factor is introduced, aimed at relaxing the boundary condition when the boundary is unclear or blurred. Secondly, in order to combine visual contents of an image, which reflects human visual response directly, a simple and general model is introduced to endow speed function with more variability and better performance. Promising experimental results in MR images are shown to demonstrate the potentials of our approach.