CT影像中毛玻璃型肺结节提取方法的研究  

GGO lung nodule extraction methods of CT images

作　　者：唐思源[1] 苗玥[1] 杨敏[1] 

机构地区：[1]包头医学院计算机科学与技术系,内蒙古包头014040

出　　处：《电子技术应用》2018年第2期109-114,共6页Application of Electronic Technique

基　　金：内蒙古自治区自然科学基金项目(2016MS0601);包头医学院科学研究基金项目(BYJJ-QM 201637)

摘　　要：毛玻璃(GGO)型肺结节目前研究的比较少,但其恶性可能性也比较大。针对GGO型结节的对比度较血管和实体型结节的对比度低,直接使用阈值法不能很好地提取GGO型肺结节,直接使用多尺度圆点滤波器不能提取非球形形状的GGO型肺结节的问题,提出了一种基于形状特征和滤波器增强的阈值法相结合的方法来提取GGO型结节。首先使用形状指数方法提取具有全部或部分球形结构的GGO型结节,然后对剔除了血管、球形结构结节的图像应用滤波增强法来拉伸毛玻璃结节和肺实质的对比度,利用阈值方法提取不具有球形结构的毛玻璃结节。为了提高检测的准确率,对肺结节的特征进行提取与分类,并选择最佳特征组合,放入支持向量机分类器提取更精确的肺结节,并对算法进行评估和对比。实验结果表明,该方法能有效降低GGO型肺结节的漏检率,提高检测的敏感性、特异性,优于现有的两种方法。The research of GGO lung nodule is rarely, but it is more possibility of malignant lesions. The direct use of the threshold method is not extract GGO lung nodules of good way due to the contrast of the GGO nodules is lower than the contrast of the blood vessels and the physical nodules, the non-spherical GGO lung nodules can＇t be extracted that it used multi-scale dot filter of directly. Therefore, the method detection of GGO nodules is proposed based on shape feature and enhance filter. First of all, shape index method is used to extract all or part of spherical structure GGO nodules, then the spheres of nodules and the vessels are removed of image, and enhancement filter is used to stretch the contrast between GGO nodule and lung parenehyma, then threshold-based approach is used to extract GGO candidate nodules. Accuracy of the test is improved, the characteristics of pul- monary nodules are extracted and selected, the optimal combination of features are trained by SVM classifier and extract pulmonary nodules, algorithms are evaluated and compared. Simulation results show that it is better than the existing two methods ,the method can reduce false detection rate, and the sensitivity and specificity are improved.

关 键 词：毛玻璃型肺结节 形状特征 滤波器增强 支持向量机分类器 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]