CT图像肺结节的毛刺检测与量化评估  

Detection and quantitative evaluation of lung nodule spiculation in CT images

作　　者：邢谦谦 刘哲星[1] 林炳权[2] 钱俊[1] 曹蕾[1] 

机构地区：[1]南方医科大学生物医学工程学院,广州510515 [2]南方医院影像中心,广州510515

出　　处：《计算机应用》2014年第12期3599-3604,共6页journal of Computer Applications

基　　金：国家自然科学基金青年基金资助项目(61102114);国家自然科学基金资助项目(61372063);广东省教育部产学研项目(2011B090400037)

摘　　要：为准确检测并量化评估毛刺征,提出一种CT图像肺结节的毛刺检测与量化评估方法。首先利用区域生长算法与水平集方法结合进行结节主体的准确分割;而后利用线性滤波模板提取结节主体周边区域的毛刺;最后引入毛刺水平指数作为毛刺特征的量化指标。在此基础上对结节有无毛刺进行分类,并与肺部图像数据库联盟(LIDC)的量化评级进行一致性和相关性分析。实验结果表明,该方法可以有效地检测并定量描述CT图像肺结节的毛刺征。A new method was proposed to accurately detect and quantitatively evaluate the lung nodule spiculation. First, the region growing method followed by level set method was used to accurately segment the main part of the lung nodule. Then, spiculated lines connected to the nodule boundary were extracted using a line detector in polar coordinates system. Finally, spiculation index was introduced as the quantitative measurement of spiculation features, which was then used as a criteria for distinguishing between spiculated and non-spiculated nodules. The consistency and correlation of spiculation index of the method and Lung Image Database Consortium （LIDC） were evaluated in detail. The experimental results show that the proposed method can effectively detect and quantitatively describe the lung nodule spiculation in CT images.

关 键 词：肺结节 毛刺 图像分割 毛刺水平指数 计算机辅助诊断 

分 类 号：N391.41[自然科学总论]