基于CT图像的肺结节检测与识别  

Detection and recognition of pulmonary nodules based on CT images

作　　者：唐思源[1] 刘燕茹 杨敏[1] 徐瑞英 TANG Siyuan;LIU Yanru;YANG Min;XU Ruiying(Department of Computer Science and Technology,Baotou Medical College of Inner Mongolia University of Science & Technology,Baotou 014040,China;Department of Medical Technology,Baotou Medical College of Inner Mongolia University of Science & Technology,Baotou 014040,China)

机构地区：[1]内蒙古科技大学包头医学院计算机科学与技术系,内蒙古包头014040 [2]内蒙古科技大学包头医学院医学技术系,内蒙古包头014040

出　　处：《中国医学物理学杂志》2019年第7期800-807,共8页Chinese Journal of Medical Physics

基　　金：内蒙古自治区自然科学基金(2016MS0601);包头医学院科学研究基金(BYJJ-QM 201637);包头医学院大学生创新创业训练计划项目(BYDCXL-201922)

摘　　要：目的:将肺结节从含有背景、噪声的胸腔区域里检测并识别出来。方法:首先,将DICOM格式的医学图像转换成JPG图像后,应用区域生长法分割出肺实质区域,去掉肺区外的干扰信息。然后,利用多尺度高斯滤波器增强图像后,应用模糊C均值聚类算法提取肺结节感兴趣区域。最后,对肺结节特征进行提取及归一化处理,应用支持向量机分类器识别并标记出肺结节。结果:在随机抽取的120例图像中,检测肺结节的准确率达到92.3%,分类识别肺结节的准确率达到95.6%。实验结果表明,本文方法有效地排除了交叉状和条形状血管等干扰,实现了肺结节的精确检测和识别。结论:本方法在保证检测和识别出正确结节的前提下,降低了误判率,算法也得到了较好的收敛。Objective To detect and identify pulmonary nodules from thoracic regions with background and noise.Methods After DICOM-format medical images were converted into JPG images,region growing method was applied to segment lung parenchyma and remove interference information outside lung area.Subsequently,multi-scale Gaussian filter was used to enhance images, and fuzzy C-means clustering algorithm was applied to extract regions of interest of pulmonary nodules.Finally,the features of pulmonary nodules were extracted and normalized,and the pulmonary nodules were identified and marked with support vector machine classifier.Results For the random sample of 120 images,the detection rate of pulmonary nodules reached 92.3% and the accuracy rate of the classification and recognition of pulmonary nodules was up to 95.6%.The experimental results revealed that using the proposed method could effectively eliminate the disturbances from crossing- and strip-shaped blood vessels and other disturbances,realizing an accurate detection and recognition of pulmonary nodules.Conclusion Using the proposed method can not only achieve an accurate detection and recognition of pulmonary nodules,but also reduce misjudgment rate.Moreover, the proposed algorithm has a better convergence.

关 键 词：肺结节 CT图像 区域生长法 多尺度高斯滤波器 模糊C均值聚类算法 支持向量机分类器 

分 类 号：R318[医药卫生—生物医学工程] TP391.41[医药卫生—基础医学]