A Fast Automatic Segmentation Method Based on Improved Hammer Elastic Registration

Instrument segmentation in ultrasound image plays a significant role in registration between surgical robot system and ultrasound assistant system. An automatic pre-extraction can provide a relatively accurate initial position for further refined segmentation. In this paper, we build up a platform to capture ultrasound images. To eliminate the error introduced by deformation in ultrasound images with and without instrument, an improved Hammer algorithm is firstly exerted, which uses a new-defined attribute vector to describe the importance of each pixel in image, and uses mesh as well as B-spline interpolation to accomplish the whole image elastic registration. Next, the difference is made between deformed image and image with instrument, so the contour of instrument can be obtained. The method can extract the contour of instrument tip quickly and automatically, which can avoid errors lead by initial position selection and is suitable to be used as initial position for further segmentation.