基于混合成像的孤立性肺结节良恶性预测模型的研究
发布时间：2018-08-23 08:42
【摘要】：随着饮食环境问题的不断加重，近年来，肺部疾病的发病率也呈不断上升的事态，已然成为了当前影响人类生活质量甚至生命的大敌因此，如何能够在病变早期就能准确的诊断出病变良恶性质，成为我们能够大大减低肺癌发病率的最有效的手段，同时也成了当前的研究热点之一在病变的最早期，肺部疾病通常在影像学上表现为肺结节，同时，良恶性结节在影像学的征象上也有极大的差异性本文中主要是基于孤立性肺结节良恶性不同的影像学征象进而实现早期良恶性鉴别 基于PET/CT的混合成像技术对肺癌进行医学诊断，在考虑肺结节的临床征象的基础上，充分结合了肺结节的PET征象和CT征象，较好的克服了单一图像对结节诊断信息不足的缺点目前，对早期肺结节良恶性的判别仍是依赖医师的阅片经验，而且所依据的判别特征不能量化，难免会出现漏诊误诊的情况为了能够尽可能的在前期减少由于主观因素而造成的漏诊误诊的现象，需要对结节各个征象进行量化，依据肺结节各个影像学征象之间的相关性建立预测肺结节良恶性的数学模型分析各个征象之间的相关度以及与肺结节良恶性的关系，从而构建一个能够预测结节良恶性的模型 本文立足于对孤立性肺结节进行良恶性预测的这一课题，主要的研究工作从以下几个方面展开： 1．利用一种改进的支持向量机——双向隶属度的模糊支持向量机的方法对孤立性肺结节良恶性进行分类本文的终极目标是实现对孤立性肺结节进行良恶性分类，利用传统的支持向量机对肺结节进行良恶性分类时，认为所有的样本对获得最优超分类面的贡献是相同的，没有考虑到样本之间的相关性对分类面的影响本文基于双向隶属度的模糊支持向量机的分类方法在对肺结节进行良恶性分类过程中充分考虑了不同样本点对分类结果的贡献，基于当前影像学中对肺结节良恶性进行诊断的比较成熟的规则，并充分考虑结节的CT PET图像征象及病变的临床征象，实现对孤立性肺结节良恶性的准确分类； 2．构建一个能充分考虑肺结节的PET和CT征象的良恶性预测的模型对所提取的结节的CT和PET征象进行量化，通过单因素分析法分析每个结节的征象与良恶性之间的关系，筛选出具有显著相关性的征象，，然后再基于所筛选出的各个因素构建能够预测结节良恶性的回归方程 最后，本论文还对所涉及的方法进行了实验，并验证各个方法的有效性，实验结果参数证明，本文的方法在鉴别肺结节良恶性方面具有较好的性能，在保证准确率的同时降低了检测的漏诊率，也在一定程度上体现了方法的泛化性
[Abstract]:With the increasing problem of diet and environment, in recent years, the incidence of lung diseases has been on the rise, which has become a major enemy affecting the quality of human life and even life. How to accurately diagnose the benign and malignant nature of the disease at the early stage of the disease has become the most effective means for us to reduce the incidence of lung cancer greatly. At the same time, it has also become one of the current research hotspots in the early stage of the disease. Lung disease is usually shown as a pulmonary nodule on imaging, and at the same time, There is also a great difference in imaging signs between benign and malignant nodules. In this paper, the imaging features of benign and malignant solitary pulmonary nodules are mainly based on the different imaging signs, so as to realize the early differentiation of benign and malignant nodules based on mixed imaging based on PET/CT. Technology for medical diagnosis of lung cancer, On the basis of considering the clinical features of pulmonary nodules, the PET and CT signs of pulmonary nodules are fully combined to overcome the shortcomings of single image in diagnosis of nodules. The diagnosis of benign and malignant pulmonary nodules is still dependent on the physician's experience in film reading, and the discriminant characteristics on which they are based cannot be quantified. In order to reduce the misdiagnosis caused by subjective factors in the early stage, it is necessary to quantify the various signs of the nodules. According to the correlation between various imaging signs of pulmonary nodules, a mathematical model was established to predict the benign and malignant pulmonary nodules. So as to build a model that can predict benign and malignant nodules, this paper is based on the subject of predicting benign and malignant solitary pulmonary nodules. The main research works are as follows: 1. Using an improved support vector machine-fuzzy support vector machine with bidirectional membership to classify the benign and malignant solitary pulmonary nodules. The ultimate goal is to achieve benign and malignant classification of solitary pulmonary nodules When using traditional support vector machine to classify lung nodules from benign and malignant, it is considered that all samples have the same contribution to obtaining the optimal super-classification surface. In this paper, fuzzy support vector machine based on bidirectional membership degree is used to classify lung nodules. In the process of classification of benign and malignant nodules, the contribution of different sample points to classification results is fully considered. Based on the mature rules for the diagnosis of benign and malignant pulmonary nodules in current imaging, and taking into account the CT PET imaging signs and the clinical signs of lesions, the accurate classification of benign and malignant solitary pulmonary nodules can be realized. 2.Construct a model that can fully consider the PET and CT signs of pulmonary nodules, and quantify the CT and PET signs of the extracted nodules. Single factor analysis was used to analyze the relationship between the signs of each nodule and the benign and malignant ones, and the signs with significant correlation were screened out. Then, based on the selected factors, the regression equation is constructed to predict the benign and malignant nodules. Finally, the methods involved are tested, and the validity of the methods is verified. The method in this paper has good performance in differentiating benign and malignant pulmonary nodules. It not only ensures the accuracy but also reduces the missed diagnosis rate of detection. It also reflects the generalization of the method to a certain extent.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP18;O212.1
本文编号：2198518
[Abstract]:With the increasing problem of diet and environment, in recent years, the incidence of lung diseases has been on the rise, which has become a major enemy affecting the quality of human life and even life. How to accurately diagnose the benign and malignant nature of the disease at the early stage of the disease has become the most effective means for us to reduce the incidence of lung cancer greatly. At the same time, it has also become one of the current research hotspots in the early stage of the disease. Lung disease is usually shown as a pulmonary nodule on imaging, and at the same time, There is also a great difference in imaging signs between benign and malignant nodules. In this paper, the imaging features of benign and malignant solitary pulmonary nodules are mainly based on the different imaging signs, so as to realize the early differentiation of benign and malignant nodules based on mixed imaging based on PET/CT. Technology for medical diagnosis of lung cancer, On the basis of considering the clinical features of pulmonary nodules, the PET and CT signs of pulmonary nodules are fully combined to overcome the shortcomings of single image in diagnosis of nodules. The diagnosis of benign and malignant pulmonary nodules is still dependent on the physician's experience in film reading, and the discriminant characteristics on which they are based cannot be quantified. In order to reduce the misdiagnosis caused by subjective factors in the early stage, it is necessary to quantify the various signs of the nodules. According to the correlation between various imaging signs of pulmonary nodules, a mathematical model was established to predict the benign and malignant pulmonary nodules. So as to build a model that can predict benign and malignant nodules, this paper is based on the subject of predicting benign and malignant solitary pulmonary nodules. The main research works are as follows: 1. Using an improved support vector machine-fuzzy support vector machine with bidirectional membership to classify the benign and malignant solitary pulmonary nodules. The ultimate goal is to achieve benign and malignant classification of solitary pulmonary nodules When using traditional support vector machine to classify lung nodules from benign and malignant, it is considered that all samples have the same contribution to obtaining the optimal super-classification surface. In this paper, fuzzy support vector machine based on bidirectional membership degree is used to classify lung nodules. In the process of classification of benign and malignant nodules, the contribution of different sample points to classification results is fully considered. Based on the mature rules for the diagnosis of benign and malignant pulmonary nodules in current imaging, and taking into account the CT PET imaging signs and the clinical signs of lesions, the accurate classification of benign and malignant solitary pulmonary nodules can be realized. 2.Construct a model that can fully consider the PET and CT signs of pulmonary nodules, and quantify the CT and PET signs of the extracted nodules. Single factor analysis was used to analyze the relationship between the signs of each nodule and the benign and malignant ones, and the signs with significant correlation were screened out. Then, based on the selected factors, the regression equation is constructed to predict the benign and malignant nodules. Finally, the methods involved are tested, and the validity of the methods is verified. The method in this paper has good performance in differentiating benign and malignant pulmonary nodules. It not only ensures the accuracy but also reduces the missed diagnosis rate of detection. It also reflects the generalization of the method to a certain extent.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP18;O212.1
【参考文献】
相关期刊论文 前4条
1 钱可宝;张勇;黄云超;;孤立性肺结节的外科治疗分析[J];中国肺癌杂志;2007年04期
2 董莘;;孤立性肺结节影像学诊断和处理[J];沈阳部队医药;2006年06期
3 王利伟;王自正;顾建平;王丽萍;冯敏;王绍娟;;孤立性肺结节的CT诊断[J];医学影像学杂志;2006年08期
4 王锋;王藏海;朱斌;;孤立性肺结节的综合CT诊断[J];河北北方学院学报;2006年06期
本文编号：2198518
本文链接：https://www.wllwen.com/kejilunwen/yysx/2198518.html
