肺实质CT图像细化分割

发布时间：2018-03-06 05:38

本文选题：肺实质分割　切入点：超像素　出处：《中国图象图形学报》2017年01期 　论文类型：期刊论文

【摘要】：目的由于肺部CT图像中各组织结构复杂、灰度分布不均匀,造成肺实质部分难以准确分割和提取。为了提高肺实质分割的准确率,本文提出了一种基于超像素的细化分割与模糊C均值聚类相结合的自动分割算法。方法该算法充分利用肺部CT图像的灰度、纹理特征,同时为了正确标记超像素的分类,引入一种空间邻域信息来增强空间约束进而有效地解决灰度不均匀的问题,它能够对肺实质进行分割并除去其周围的主血管,然后利用形态学知识去除肺部的分支血管。结果在临床患有四类疾病的患者CT图像数据集上采用改进的图像特征,使得肺实质分割的准确率提高了0.8%。同时,算法准确率提高到99.46%。结论实验结果表明,本文算法能够实现肺部CT图像肺实质的自动细化分割,结果准确适用。该算法鲁棒性好、速度快,是一种精确有效的自动肺实质分割方法。
[Abstract]:Objective to improve the accuracy of lung parenchyma segmentation, it is difficult to segment and extract the lung parenchyma due to the complicated tissue structure and uneven gray distribution in lung CT images. In this paper, an automatic segmentation algorithm based on super-pixel thinning and fuzzy C-means clustering is proposed. A spatial neighborhood information is introduced to enhance the spatial constraint and solve the problem of uneven gray scale effectively. It can segment the lung parenchyma and remove the main blood vessels around it. Then the branch blood vessels of the lung were removed by morphological knowledge. Results the improved image features were used in the CT image data set of patients with four kinds of diseases, which increased the accuracy of lung parenchyma segmentation by 0.8%. The experimental results show that the algorithm can realize automatic thinning and segmentation of lung parenchyma in lung CT images, and the results are accurate and applicable. The algorithm is robust and fast. It is an accurate and effective method for automatic pulmonary parenchyma segmentation.
【作者单位】：山东中医药大学理工学院;山东大学计算机科学与技术学院;
【基金】：国家自然科学基金项目(U1201258,61572300) 山东省自然科学基金项目(ZR2015FM010) 山东高等学校科技计划项目(J15LN20) 山东省医药卫生科技发展计划项目(2016WSO577) 山东省中医药科技发展计划项目(2015-026)~~
【分类号】：TP391.41

【相似文献】