Tunneling descent level set segmentation of ultrasound images

摘要: The presence of speckle in ultrasound images causes many spurious local minima the energy function active contours. These trap segmentation prematurely under gradient descent and cause algorithm to fail. This paper presents a substantially new reformulation Tunneling Descent, which is deterministic technique escape from unwanted minima. In formulation, evolving curve represented by level sets, evolution strategy obtained as sequence constrained minimizations. The used segment endocardium 115 short axis cardiac images. All segmentations are achieved without tweaking or numerical parameters. Experimental evaluation results shows that overcomes multiple give considerably more accurate than conventional techniques.