一种基于CI特征的3-域均值平移聚类肺结节分割算法(英文)  

Pulmonary nodule segmentation algorithm based on three-domain mean shift clustering

作　　者：聂生东[1] 李立鸿[2] 陈兆学[1] 

机构地区：[1]上海理工大学医疗器械与食品学院,上海200093 [2]上海交通大学医学图像处理与模式识别研究所,上海200240

出　　处：《华东师范大学学报（自然科学版）》2008年第1期60-67,共8页Journal of East China Normal University(Natural Science)

基　　金：上海市教委重点项目(06ZZ33);上海市重点学科资助项目(P0502);上海高校选拔培养优秀青年教师科研专项基金(358536)

摘　　要：提出了一种有效的分割CT图像中肺结节的新算法.该算法采用均值平移(mean shift)算法和基于CI(Convergence Index,CI)特征,共由三个步骤组成:(1)计算感兴趣区内的所有像素的CI特征;(2)把CI特征与像素的灰度值和空间位置信息结合在一起,形成3-域特征向量集;(3)利用均值平移聚类算法对特征向量集进行聚类.由于该算法能有效分析多高斯模型描述的包括实质性结节和亚实质性结节在内的所有结节,因此.可应用于CT图像中任何含有结节的用户感兴趣区域.实验结果证明,该方法能更精确地分割出不同类型的结节.A novel and more effective algorithm used for segmenting pulmonary nodules in CT images was presented. The algorithm is based on mean shift clustering method and CI （Convergence Index） features, which can represent the multiple Gaussian model of pulmonary nodules both for solid and sub-solid, substantially. The algorithm has three steps： （1） calculating the CI feature of every pixel in the region of interest （ROD ; （2） combining the CI feature with the intensity range and the spatial position of each pixel to form a feature vector set; （3） grouping all feature vector sets into different cluster with mean shift clustering algorithm. Owing to our algorithm can represents the multiple Gaussian model both for solid and sub-solid nodules, it can be used in any interesting nodule regions, especially suitable for the segmentation of sub-solid nodules. Experiments demonstrated that our algorithm can figure out the outline of pulmonary nodules of different forms more precisely.

关 键 词：CT图像 结节分割 实质性结节 亚实质性结节 CI特征 均值平移算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TH701[自动化与计算机技术—计算机科学与技术]