一种基于蚁群算法特征建模与分类的肺结节检测方法  

Pulmonary Nodule Detection Approach of Feature Modeling and Classification Based on Ant Colony Algorithm

作　　者：郭薇[1] 魏颖[1] 周翰逊 薛定宇[1] 

机构地区：[1]东北大学信息科学与工程学院 [2]辽宁大学信息科学与技术学院

出　　处：《系统仿真学报》2010年第5期1155-1158,1163,共5页Journal of System Simulation

基　　金：国家自然科学基金(60671050);辽宁省自然科学基金(20052021)

摘　　要：根据肺结节的特征模型对感兴趣区域(ROI)进行分类来检测肺结节。首先,使用蚁群算法构造初步的规则集合,根据条件元素的相关度对规则进行修改确定最终的规则集合。然后,对规则集合中的特征进行建模,定义特征变量表示选定特征。最后,根据ROI的这些特征变量值以及总体特性,利用蚁群算法对ROI进行分类。分类中根据蚂蚁尸体堆积模型确定分类中心,采用蚂蚁觅食模型进行分类。仿真结果表明,该算法检测肺结节具有较高的敏感性,对医生后续的肺癌诊断有指导意义。According to feature models of pulmonary nodules,regions of interest （ROI） were classified to detect pulmonary nodules.Firstly,an ant colony algorithm was utilized to form the primary rule set,and then the primary rule set was modified to confirm the final rule set based on the correlation degrees of the condition elements.Secondly,the features in the final rule were modeled,and the features variables were defined to represent the features in the final rule set.Thirdly,an ant colony algorithm was used to classify ROI based on each feature variable value and population characteristic of ROI.The center of classification was determined by the model that imitated the aggregation of dead bodies by ants,and the model of ants looking for food was used to classify.The experiment results indicate that this approach for detecting pulmonary nodules has a high sensitivity,and it can provide doctors with helpful information for diagnosis of pulmonary cancer.

关 键 词：分类规则 特征提取 特征建模 蚂蚁尸体堆积模型 蚂蚁觅食模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]