基于自适应区域生长的乳腺肿块分割方法  

Segmentation of breast masses using adaptive region growing

作　　者：杨斌[1] 宋立新[1] 

机构地区：[1]哈尔滨理工大学电气与电子工程学院,哈尔滨150080

出　　处：《计算机工程与应用》2014年第20期171-175,210,共6页Computer Engineering and Applications

基　　金：黑龙江省自然科学基金(No.F200912);哈尔滨创新人才基金(No.2010RFXXS026)

摘　　要：乳腺X图像中肿块特征的复杂多变,给肿块的分割带来了很大困难,区域生长为肿块分割提供了一种比较可靠的方法。传统的区域生长由于生长次数和准则比较单一,就会出现较多的过生长和欠生长,从而影响其分割精度和可靠性,针对这一问题,提出了一种利用自适应区域生长对乳腺肿块进行分割的方法。对肿块感兴趣区域进行背景去除和领域抑制得到预处理后的图像,利用预处理后图像各像素个数确定区域生长的种子点,再利用肿块图像的梯度分布及变化趋势确定自适应区域生长是否过边缘,从而确定最佳生长准则。实验结果表明,相对于三层地形分割算法及模型分割算法,自适应区域生长算法分割得更准确、可靠。Since there are a lot of complex and changing characteristics of mass in mammography with great difficulty in mass segmentation, region growing becomes a reliable method to accomplish it. An adaptive region growing method for mass segmentation is presented so as to improve its precision and reliability and reduce the over-growing and lack-growing when dealing with different images in one principle. Background removing and region suppression are used to preprocess the Region Of Interest（ROI）of mass, and then it uses the number of image pixels to determine the seed point for region growing, and determines whether the adaptive region growing is out of edge through the gradient distribution and tends of mass ROI in order to obtain the best growth criteria. The experimental results show that the adaptive region growing algo-rithm for segmentation compared to the three-terrain segmentation algorithm and model segmentation algorithm is more accurate and reliable.

关 键 词：肿块分割 图像预处理 梯度 自适应区域生长 

分 类 号：TN911.73[电子电信—通信与信息系统]