基于局部阈值和聚类中心迭代的肺结节检测算法  

Pulmonary Nodules Detection Algorithm Based on Local Threshold and Iterative of Clustering Center

作　　者：陈侃[1] 李彬[1] 田联房[1] 

机构地区：[1]华南理工大学自动化科学与工程学院,广州510640

出　　处：《计算机科学》2012年第2期302-304,共3页Computer Science

基　　金：广东省教育部产学研结合项目(2009B090300057);教育部博士点基金资助项目(200805610018);广州市番禺区科技攻关项目(2009-Z-108-1);华南理工大学中央高校基本科研业务费(2009ZM0077)资助

摘　　要：肺部疾病通常以肺结节的形式表现出来。为了对肺部疾病进行诊断治疗,需要对肺结节进行准确的检测。提出了基于局部阈值和聚类中心迭代的肺结节检测算法。首先,对肺实质图像采用局部阈值算法,提取感兴趣区域(ROIs),并且计算ROIs的形态特征、灰度特征和纹理特征;其次,结合规则、聚类中心迭代和欧式距离,对ROIs进行分类。实验结果表明,所提算法能够较好地检测出孤立性结节、低对比度结节和粘连肺壁结节。Lung disease is usually showed in the form of pulmonary nodules.In order to diagnose lung disease,pulmonary nodules must be detected accuratly.This paper proposed pulmonary nodules detection algorithm based on local threshold and iterative of clustering center.First,local algorithm was proposed for extracting interested region（ROIs） in lung parenchyma image.Morphological features,gray features and texture features were calculated in interested region（ROIs）.Second,rule,iterative algorithm of clustering center and Euclidean distance were used to classing ROIs.Experiment results show that the proposed algorithm can better detect isolation nodules,low contrast nodules and nodules which adhere to lung wall.

关 键 词：肺结节 局部阈值 聚类中心迭代 欧式距离 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]