离散轮廓点集法提取超声图像前列腺边缘

Prostate Ultrasound Image Boundary Extraction Using Discrete Contour Point Set

提出一种基于离散轮廓点集的超声图像前列腺边缘提取方法,将经直肠超声图像中的腺体准确分割出来.该方法根据前列腺的轮廓特征对候选轮廓点进行多级筛选,再除去由于各种干扰产生的非边缘点,得到位于实际边缘的离散轮廓点集,由此确定粗糙的腺体边缘.随后利用一种快速水平集的曲线演化方法对初始边缘进行小范围调整,使其更接近实际的腺体轮廓.对临床应用中的经直肠超声图像进行边缘提取实验,结果表明该方法能克服腺体外部和内部区域灰度不均一以及边缘模糊等问题,提取的前列腺边缘位置准确,轮廓完整.

A boundary extraction method based on the discrete contour point set（DCPS） is proposed for transrectal ultrasound（TRUS） prostate images.Several features of the prostate boundary are used to search candidate contour points,and a point set is picked up from them as the actual boundary by eliminating nonboundary candidate points.A coarse prostate boundary is defined by the DCPS and adjusted in a small scale to get a closer prostate boundary using a curve evolution method with a fast level set.Results of experiments for a series of TRUS images show that the proposed method can effectively overcome problems of inhomogeneous intensity distributions inside and outside the prostate,and blur or missing edges,to extract an accurate and fully-formed boundary of the prostate.