基于多维图像信息的肺结节良恶性鉴别方法  

Discrimination of benign and malignant pulmonary nodules based on multi-dimensional features

作　　者：郭薇[1] 王柳[1] 肖男[1] 张国栋[1] 

机构地区：[1]沈阳航空航天大学计算机学院,沈阳110136

出　　处：《沈阳航空航天大学学报》2016年第3期67-72,共6页Journal of Shenyang Aerospace University

基　　金：国家自然科学基金(项目编号:61402298);国家自然科学基金(项目编号:61373088);航空科学基金(项目编号:2013ZE54025)

摘　　要：CT图像中肺结节的良恶性鉴别是肺癌计算机辅助诊断研究中的关键。针对这一问题,提出了一种基于图像多维信息的肺结节良恶性鉴别方法。该方法基于三维肺结节的二维表示来进行影像学征象建模,并计算模型特征值。然后,提取结节的形状及纹理特征。最后,根据提取的所有特征,利用模糊C均值分类器对肺结节的良恶性进行鉴别。针对130组结节数据进行实验,肺结节良恶性鉴别的敏感性及假阳率分别达到87.58%及9.52%。实验结果表明,该文提出的方法能有效地区分肺结节的良恶性,辅助医生进行临床诊断。The discrimination of benign and malignant pulmonary nodules in CT image is the key of computer-aided diagnosis( CAD) for lung cancer. For this purpose,we proposed a method to identify the benign and malignant of pulmonary nodules based on multi-dimensional features. Firstly,the models of CT signs for 3D pulmonary nodules are established based on 2D representation,and the values of the models were calculated. Then,the shape and texture features of the nodules were fetched. Finally,the fuzzy C-means classifier was employed to discriminate the benign and malignant pulmonary nodules. The sensibility and specificity of the method are 87. 58% and 9. 52% respectively. It shows the proposed method can discriminate the benign and malignant nodules well and help doctors in the clinical diagnosis.

关 键 词：计算机辅助诊断 分叶征 毛刺征 模糊C均值聚类 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]