结合呼出气、影像及风险因素的肺癌早期诊断模型研究

发布时间：2018-01-16 00:27

本文关键词：结合呼出气、影像及风险因素的肺癌早期诊断模型研究　出处：《浙江大学》2016年硕士论文　论文类型：学位论文

【摘要】：肺癌作为全球最常见和死亡率最高的恶性肿瘤,已经是全人类的公敌。而近年来肺癌在我国的发病率呈现了快速增长的趋势。但由于早期肺癌相应症状不易识别,且易转移等天然特性,导致肺癌治疗窗口期短、治愈率低。因此早期诊断显得尤为重要,它可以延长患者的治疗窗口从而提高患者的生存率。本研究在呼出气体挥发性有机物(VOCs)肺癌标识的研究基础上,结合临床的影像学判断以及肺癌风险因素,研究了相应的早期诊断模型,主要内容包括：1)检测硬件平台及流程的设计与实现：采用采气仪—热脱附仪—气相色谱—质谱联用的呼出气检测平台,检测分析了149例正常健康人、27例肺良性疾病患者以及60例肺癌患者呼出气体中vocs的表达。设计了肺癌风险因素的评估问卷,并基于问卷信息建立了肺癌风险诊断数据库。对参与研究的受试者的影像学信息进行分析,并提取影像学分类特征完善了风险诊断数据库。本论文还对平台的呼出气采集过程,样本前处理过程,样本物流过程及检测过程进行了质量控制研究。2)数据分析处理及交互的软件设计与实现：完成了vocs的数据处理分析软件,对所有呼出气样本进行去背景处理并实现呼出气数据采集分析的批量处理,和数据导入。同时为了方便参与研究的受试者查询其实验结果,延伸性的设计了一个移动端的数据交互应用。3)肺癌人群风险因素及呼出气标识物的相关性研究：分析了呼出气中VOCs影像学、吸烟行为、性别、年龄等风险因素之间的相关性,并分别找到13种、VOCs与影像学显著相关,3种VOCs与“吸烟情况”显著相关,2种VOCs与“吸烟年数”显著相关,1种VOCs与“吸烟方式”显著相关,10种VOCs与“平均每天吸烟数”显著相关,3种VOCs与性别显著相关,8种VOCs与“年龄”显著相关。4)基于相关性研究的结果构建并优化肺癌早期诊断模型：在分析单从影像学判断肺癌的敏感性和特异性后,联合呼出气中筛选出的31种与肺癌高显著相关VOCs和影像学信息、吸烟行为、性别、年龄等分别采用Logistic回归模型、BP神经网络模型以及Fisher判别模型建立肺癌早期诊断模型,结果显示相比较于单影像学判断,多因素诊断模型的确能有效的提高肺癌诊断率。本论文初步研究了呼出气标识物结合影像学判断及人群风险因素的诊断模型,为肺癌的早期二级预防提供了一种新的技术手段,我们希望随着这项研究的不断深入,在不久的将来这些模型能应用于肺癌早期诊断和筛查。
[Abstract]:Lung cancer is the most common malignant tumor in the world. In recent years, the incidence of lung cancer in China has shown a trend of rapid growth, but due to the early symptoms of lung cancer is not easy to identify, and easy to metastasize and other natural characteristics. As a result of lung cancer treatment window period is short, the cure rate is low, so early diagnosis is particularly important. It can prolong the treatment window of patients and improve the survival rate of patients. This study is based on the study of the identification of lung cancer by exhalating volatile organic compounds (VOCs). Combined with clinical imaging judgment and lung cancer risk factors, the corresponding early diagnosis model was studied. The main contents include the design and realization of the hardware platform and flow chart of the detection: 149 healthy people were detected and analyzed by using the exhalation detection platform of gas recovery instrument thermal desorption instrument gas chromatography-mass spectrometry. Expression of vocs in exhaled gases from 27 patients with benign lung diseases and 60 patients with lung cancer. A questionnaire was designed to assess the risk factors of lung cancer. The database of lung cancer risk diagnosis was established based on the questionnaire information, and the imaging information of the subjects involved in the study was analyzed. And the extraction of imaging classification features to improve the risk diagnosis database. This paper also on the platform of the exhalation process, sample pre-processing process. Sample logistics process and testing process of quality control research. 2) data analysis and data processing and interactive software design and implementation: completed the vocs data processing and analysis software. All exhalation samples are processed without background, and batch processing and data import of data collection and analysis of exhalation data are realized. At the same time, in order to facilitate the participants in the study to query the results of the experiment. Extended design of a mobile end of the data interaction application. 3) Lung cancer population risk factors and exhalation markers correlation study: analysis of the exhalation of VOCs imaging, smoking behavior, sex. Age and other risk factors were correlated, and 13 VOCs were found to be significantly correlated with imaging. 3 VOCs were significantly correlated with smoking status. Two kinds of VOCs were significantly correlated with "number of years of smoking" and one VOCs was significantly correlated with "smoking mode". 10 kinds of VOCs were significantly correlated with "average daily smoking number". Three kinds of VOCs were significantly correlated with sex. Eight types of VOCs were significantly correlated with age. 4) based on the results of correlation studies, early diagnosis models of lung cancer were constructed and optimized: after analyzing the sensitivity and specificity of lung cancer from imaging alone. 31 kinds of lung cancer selected from combined exhalation were highly correlated with VOCs and imaging information, smoking behavior, sex, age and so on, respectively using Logistic regression model. BP neural network model and Fisher discriminant model were used to establish the early diagnosis model of lung cancer. Multi-factor diagnosis model can effectively improve the diagnosis rate of lung cancer. This paper preliminarily studied the diagnostic model of breath out marker combined with imaging judgment and population risk factors. It is hoped that these models can be used in early diagnosis and screening of lung cancer in the near future.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：R734.2

【相似文献】