The Optimization And Application Of Lung Cancer Early Diagnosis Model With Volatile Organic Compounds In Exhaled Gas

Lung cancer,recognized as the most common primary malignancy of the lungs,is threatening human health.The symptoms of early lung cancer could take subtle forms,which contributes to the indefinite diagnosis of lung cancer,in other words,lung cancer can only be confirmed in the middle and late stages.In this condition,the therapeutic time window is shorter,while the treatment is more difficult and the cure rate is lower.Thus,rapid and accurate early diagnosis of lung cancer can extend the therapeutic time window,thereby reduce the mortality of lung cancer.This thesis optimizes the model of early diagnosis on lung cancer through the exhaled volatile organic compounds(VOCs)and discusses the feasibility of the model in clinical application,the contents including:(1)Standardizing the process of collecting exhaled VOCs.The collection process of experimental data requires a gas collector for enriched exhaled breath,a thermal desorption device for the sample desorption,and a gas chromatograph-mass spectrometer for the separation of the test sample components.The thesis specifies the instrument parameter settings,instrument operation steps and the procedures for the process.In the thesis,we also electronize the questionnaires which collecte the basic information of the volunteers and make data collection automate and program.(2)Designing the online database of lung cancer.We achieve the entry and query of the lung cancer data online,providing convenience for data storage and management.Lung cancer database contains the basic information,imaging diagnostic information as well as the exhaled VOCs data of the volunteers.(3)Determining the VOCs that can be used to optimize lung cancer diagnosis models.We confirm seven kinds of VOCs associated with smoking status,seven with smoking intensity,eleven with smoking years and twenty-seven with smoking depth.The distribution of VOCs in the human population is analyzed,and the feasibility of smoking-related VOCs in lung cancer diagnosis models is also discussed.(4)Optimizing the model of early diagnosis on lung cancer and evaluating the value of the model in clinical application.In this thesis,smoking-related VOCs are used to optimize the diagnostic model of lung cancer based on BP neural network,the Fisher linear discriminant and logistic regression to improve the sensitivity,specificity and overall accuracy of the model.The validated set is used to predict the optimized model,the logistic regression models performs the best,which has the potential in clinical application.The sensitivity of the model is 96.5%,the sensitivity of the validation set is 93.8%.