基于CT影像的肺部气管树分割算法的研究
发布时间：2019-04-24 04:51
【摘要】：背景：在现代社会中,随着工业化的发展,导致空气污染加重,人体肺部疾病发病率高,类型众多,如肺癌,气管炎,肺气肿,哮喘等都会严重危害人体健康。随着多排螺旋CT的发展,医生可以通过一次扫描便获得患者的影像信息,并结合先进的影像处理技术对疾病进行分析诊断。基于上述优点,CT成像并认为是诊断肺部疾病的金标准。但由于气管树的结构复杂,对比度差,并且容易受到噪声、容积效应的影响,如何准确的在CT数据中提取完整的气管树结构仍面临挑战。 目的：基于人体气管树的形态学结构,提出一种全新的方法来提取气管树结构,并为临床诊断提供参考依据。材料和方法：本课题采用的所有CT数据全部来自辽宁省沈阳市盛京医院,所有的CT图像的重建矩阵大小皆为512*512,机器型号包括飞利浦Brilliance64和东芝。本算法首先基于区域生长,波传递算法,形态优化和填洞操作来提取主气管。再提取肺区其余部位的低CT值区域,并使用一些形态学手段来得到肺区内的管状结构特征组织,之后筛选正确的遗漏气管并将其和主气管进行拼接,最后再通过迭代操作来得到气管分割的最佳结果。 结果：此算法已在成功应用于28个肺部CT数据,并邀请盛京医院的放射科医生对分割结果进行评估,并与人体实际支气管的解剖学特征进行对比,根据统计结果可得本算法能够分割出5级以上的人体支气管。 结论：人体气管树的分割结果能为医生提供一定的临床参考,并为气管分支分段骨架化和气管仿真内窥镜提供研究基础。
[Abstract]:Background: in modern society, with the development of industrialization, air pollution aggravates, the incidence of human lung diseases is high, many types, such as lung cancer, tracheitis, emphysema, asthma and so on, will seriously endanger human health. With the development of multi-slice spiral CT, doctors can obtain the patient's image information through one scan, and combine with advanced image processing technology to analyze and diagnose the disease. Based on these advantages, CT imaging is considered to be a gold standard for the diagnosis of pulmonary diseases. However, because the structure of trachea tree is complex, the contrast is poor, and it is easy to be affected by noise and volume effect, how to accurately extract the complete trachea tree structure from CT data is still a challenge. Aim: based on the morphological structure of human trachea tree, a new method was proposed to extract trachea tree structure and provide reference for clinical diagnosis. Materials and methods: all the CT data used in this study are from Shengjing Hospital, Shenyang, Liaoning Province. The reconstruction matrix size of all CT images is 512 / 512. The machine models include Philips Brilliance64 and Toshiba. Firstly, the algorithm is based on region growth, wave transfer algorithm, shape optimization and hole filling operation to extract the main trachea. Then the low CT value region of the rest of the lung area was extracted, and some morphological methods were used to obtain the tubular structure characteristic tissue in the lung area, and then the correct missing trachea was screened and spliced with the main trachea. Finally, the optimal results of trachea segmentation are obtained by iterative operation. Results: this algorithm has been successfully applied to 28 lung CT data. Radiologists in Shengjing Hospital are invited to evaluate the segmentation results and compare them with the anatomical characteristics of the actual bronchus in human body. According to the statistical results, the algorithm can be used to segment human bronchi at or above level 5. Conclusion: the segmentation results of human trachea tree can provide a certain clinical reference for doctors, and provide a basis for the research of segmental ossification of trachea branches and tracheal virtual endoscopy.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R816.41;TP391.41
本文编号：2464117
[Abstract]:Background: in modern society, with the development of industrialization, air pollution aggravates, the incidence of human lung diseases is high, many types, such as lung cancer, tracheitis, emphysema, asthma and so on, will seriously endanger human health. With the development of multi-slice spiral CT, doctors can obtain the patient's image information through one scan, and combine with advanced image processing technology to analyze and diagnose the disease. Based on these advantages, CT imaging is considered to be a gold standard for the diagnosis of pulmonary diseases. However, because the structure of trachea tree is complex, the contrast is poor, and it is easy to be affected by noise and volume effect, how to accurately extract the complete trachea tree structure from CT data is still a challenge. Aim: based on the morphological structure of human trachea tree, a new method was proposed to extract trachea tree structure and provide reference for clinical diagnosis. Materials and methods: all the CT data used in this study are from Shengjing Hospital, Shenyang, Liaoning Province. The reconstruction matrix size of all CT images is 512 / 512. The machine models include Philips Brilliance64 and Toshiba. Firstly, the algorithm is based on region growth, wave transfer algorithm, shape optimization and hole filling operation to extract the main trachea. Then the low CT value region of the rest of the lung area was extracted, and some morphological methods were used to obtain the tubular structure characteristic tissue in the lung area, and then the correct missing trachea was screened and spliced with the main trachea. Finally, the optimal results of trachea segmentation are obtained by iterative operation. Results: this algorithm has been successfully applied to 28 lung CT data. Radiologists in Shengjing Hospital are invited to evaluate the segmentation results and compare them with the anatomical characteristics of the actual bronchus in human body. According to the statistical results, the algorithm can be used to segment human bronchi at or above level 5. Conclusion: the segmentation results of human trachea tree can provide a certain clinical reference for doctors, and provide a basis for the research of segmental ossification of trachea branches and tracheal virtual endoscopy.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R816.41;TP391.41
【参考文献】
相关期刊论文 前4条
1 姜慧研;张晔;;基于改进的区域生长法的气管与支气管分割[J];东北大学学报(自然科学版);2009年02期
2 王胜军;付玲;康雁;刘积仁;;一种基于管状特征的冠脉中心线自动提取算法[J];东北大学学报(自然科学版);2011年01期
3 程远雄;胡罢生;郝立巍;陈思平;;一种新的三维气管树提取方法[J];中国医学物理学杂志;2011年04期
4 翟伟明;胡成文;张伟宏;牟文斌;赵雁南;贾培发;;基于动态自适应体素生长的肺部CT图像3维分割算法[J];中国图象图形学报;2005年10期
，本文编号：2464117
本文链接：https://www.wllwen.com/yixuelunwen/yundongyixue/2464117.html
最近更新
教材专著
