基于CT图像的亚实性肺结节实性成分的提取

发布时间：2018-04-10 15:22

本文选题：CT图像 + 肺结节　；参考：《天津工业大学》2017年硕士论文

【摘要】：据世界各大癌症协会报道,肺癌正逐步成为威胁人类健康的癌症之首。肺结节是肺癌早期表现之一,鉴别肺结节的良恶性在肺结节患者的整个治疗过程中占据极其重要的位置。本文提出的期望最大化法(Expectation Maximization Algorithm,EM)法能成功提取出肺结节中的实性成分,缩短了医生手动分割的时间,为医生鉴别肺结节的良恶性提供一定的参考依据。亚实性肺结节恶性率较高,实性成分的大小可以作为判断肺结节良恶性的一个参考,而目前针对CT图像中亚实性肺结节实性成分的研究相对较少。针对这个问题,本文共进行了以下几个方面的工作:分割出肺实质图像。本文在这一环节提出多次重复使用数学形态学运算将相互粘连的左右肺区分开的方法,该方法运算简单速度快。对所进行实验的20个病例中含有肺结节的120张图像,分割出肺实质的准确率为100%。在肺结节检测的问题上,本文共采用两种方法,其一为Snake活动轮廓法,其二为先使用模糊C均值聚类法检测所有疑似肺结节,由于直接得到的结果假阳性肺结节较多,本文加入了对目标形态特征的分析进一步降低假阳性率的方法,最终将肺结节分割出来。本文提出使用EM法提取肺结节实性成分,为了验证该方法的稳定性,本文采用两种方式分割肺结节,并使用EM法对肺结节一一进行实性成分的提取。得到的实性成分图像与Philips viewer软件中观察的图像做比较,探究亚实性肺结节实性成分提取的效果。通过对20个病例的实验,结果表明,本文基于CT图像采用EM法可以成功提取出亚实性肺结节中实性成分,结果与参考图像十分接近,避免了因医生手动分割而带来的不确定因素,很大程度上缩短了医生分割实性成分的时间,减少其工作负担,并成功计算出实性成分的体积占据肺结节体积的比值,为医生判定肺结节的良恶性提供重要的参考依据。
[Abstract]:Lung cancer is becoming the leading cancer threat to human health, according to the world's major cancer associations.Pulmonary nodules are one of the early manifestations of lung cancer. Differentiating benign and malignant pulmonary nodules plays an important role in the treatment of pulmonary nodules.The expectation maximization Maximization algorithm proposed in this paper can successfully extract the solid components from pulmonary nodules, shorten the time of manual segmentation by doctors, and provide a certain reference for doctors to distinguish benign and malignant pulmonary nodules.The malignant rate of subsolid pulmonary nodules is high and the size of solid components can be used as a reference for the diagnosis of benign and malignant pulmonary nodules.In order to solve this problem, we have done the following work: segmenting the lung parenchyma image.In this part, a method of dividing the conglutinated left and right lungs by repeated mathematical morphological operation is proposed, which is simple and fast.120 images of pulmonary nodules in 20 cases were used to segment pulmonary parenchyma with 100 accuracy.There are two methods to detect pulmonary nodules, one is Snake active contour method, the other is using fuzzy C-means clustering method to detect all suspected pulmonary nodules.In this paper, the method of reducing false positive rate is added to analyze the morphological characteristics of the target, and finally the pulmonary nodules are segmented.In this paper, EM method is used to extract the solid components of pulmonary nodules. In order to verify the stability of this method, two methods are used to segment pulmonary nodules, and EM method is used to extract the solid components of pulmonary nodules one by one.The solid component images obtained were compared with those observed in Philips viewer software to explore the effect of subsolid pulmonary nodule solid component extraction.The experimental results of 20 cases showed that the solid components of subsolid pulmonary nodules could be extracted successfully by using EM method based on CT images, and the results were very close to those of reference images.It avoids the uncertainty caused by the doctor's manual segmentation, greatly shortens the time for doctors to divide solid components, reduces their workload, and successfully calculates the ratio of the volume of solid components to the volume of pulmonary nodules.To provide an important reference for doctors to determine the benign and malignant pulmonary nodules.
【学位授予单位】：天津工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R734.2;TP391.41

【参考文献】