A Graph-based Segmentation Method for Breast Tumors in Ultrasound Images
This paper introduces a graph-based imagesegmentation method for detecting breast tumors in ultrasoundimages. The proposed segmentation algorithm based on theminimum spanning trees in a graph generated from an image,can automatically detect tumor regions and segment lesions inultrasound images. The algorithm for segmenting breastultrasound images consists of 3 steps, i.e. the nonlinear coherentdiffusion model for speckle reduction, the graph construction formapping the image to a graph, and the mergence of smallerregions. A pairwise region comparison predicate comparing theinter-component differences with the within componentdifferences, is used to determine whether or not two regionsshould be merged. Experimental results have shown that theproposed segmentation algorithm is simply structured, robust tonoises, highly efficient and much flexible in comparison withFuzzy C means clustering. It can successfully detect tumors andextract lesions in ultrasound images more accurately. We hopethat our method could be useful in various medical practices,providing an alternative way for ultrasound image analysis.
Suying Lee Qinghua Huang Lianwen Jin Minhua Lu Tianfu Wang
School of Electronic and Information Engineering,South China University of Technology,Guangzhou, Chi Department of Biomedical Engineering,Shenzhen University,Shenzhen, China
国际会议
成都
英文
1-4
2010-06-18（万方平台首次上网日期，不代表论文的发表时间）