The improved level set evolution for ultrasound image segmentation in the high-intensity focused ultrasound ablation therapy

Abstract An improved level set evolution model is proposed to segment the ultrasound images accurately and efficiently in this paper. At first, the multi-scale gradient vector field is employed to reduce the image noise. We assign the standard deviation as a scale to set up several scales images, and use the gradient direction information between the adjacent scales to refine the multi-scale gradient vector field, which highlights the real boundaries and reduces the noise. Besides, the divergence operator of the multi-scale gradient vector field is used to overcome the sensitivity problem of initialization and avoid boundary leakage. The experiments results demonstrate that the performance of the proposed method is very close to the ground truths obtained from manual segmentation by specialists with higher area similarity measure (ASM) and lower mean contour distance (MCD), which demonstrated that the proposed method is reliable, accurate and robust for the tumor boundary segmentation in ultrasound images.