The Value Of CT Features And Texture Analysis For Predicting EGFR Mutation Of Lung Adenocarcinoma

PART 1 Correlaion between CT features and epidermal growth factor receptor mutations in lung adenocarcinomaObjective: To investigate the correlations between computed tomography(CT)features and epidermal growth factor receptor(EGFR)mutations in lung adenocarcinoma.Methods: Imaging date of 156 patients with lung adenocarcinoma diagnosed,confirmed by pathology and EGFR gene detection in our hospital from June 2016 to June 2019,were retrospectively collected.Among them,there were 104 cases of EGFR mutant,40 males and 64 females,with an average age of(62.68±12.65)years;52 EGFR wild-type cases,38 males and 14 females,with an average age(60.33±12.05)years old.All patients were randomly divided into training set(n=110)and verification set(n=46).To collect CT imaging features,including location,outline of the tumor(lobulation sign,burr sign,clear or blurred boundary),internal signs of the tumor(ground glass density shadow,vacuole sign,cavity,air bronchial sign,calcification,bronchiectasis,bronchial stenosis),pleural indentation sign,emphysema,pulmonary vesicles,pleural effusion and enlarged lymph nodes.The quantitative data were analyzed by independent-sample T test,the qualitative data were analyzed by chi-square test or Fisher’s exact test.The variables with statistical significance in univariate analysis were analyzed by multivariate logistic regression analysis to investigate the correlations between clinical as well as CT imaging features and EGFR gene mutation in lung adenocarcinoma.Results: Univariate analysis showed that female,ground-glass opacity and no emphysema were associated with EGFR mutation,and there were significant difference between EGFR mutation group and wild group(P<0.05).There were no significant correlation between age and other CT features,include location,burr sign,vacuole sign,cavity,air bronchial sign,calcification,bronchiectasis,bronchostenosis,pleural indentation sign,pulmonary vesicle,pleural effusion and lymph node enlargement and there were no significant difference between mutation group and wild group(P>0.05).Logistic regression analysis showed that only female was an independent risk factor for EGFR gene mutation in lung adenocarcinoma,and there were no significant difference between GGO and emphysema and EGFR mutation status.Conclusion: Sex is associated with EGFR gene mutation in lung adenocarcinoma,which may be used as an independent risk factor for predicting EGFR gene mutation in lung adenocarcinoma.PART 2 Values of CT texture anaiysis in predicting epidermal growth factor receptor mutation in lung adenocarcinomaObjective: To evaluate the value of CT texture anaiysis in predicting the mutation state of epidermal growth factor receptor(EGFR)in lung adenocarcinoma based on CT plain scanning images.Methods: To Collected the CT images of the patients in the first part and imported them into ITK-SNAP software.Drew the three dimensional volume region of interest in the lung window with ITK-SNAP software.The CT texture features were extracted by AK software.The most valuable predictor subset was screened by m RMR,LASSO regression and 10-fold cross verification.Gender,ground-glass opacity and emphysema were used to construct clinical model,and selected CT texture features were used to construct imaging model,and to calculate the radsocre of each patient.The joint model was constructed by using clinical model and radsocre,and the nomogram of the joint model was drawn.The ROC curve was constructed to evalauted the efficiency of the different models,and the calibration curve was used to evaluated the consistency between the predicted value and the observed value.The clinical utility value of the models were evaluated with decision curve analysis(DCA).Results: A total of 396 CT texture features were extracted,and 9 features with the most predictive value were obtained to construct the imaging model,the training set AUC was 0.846(95%CI:0.766-0.926),the specificity and sensitivity were 0.865 and 0.726 respectively,and the verification set AUC was 0.794(95%CI:0.608-0.979),the specificity and sensitivity were 0.800 and 0.871 respectively.The clinical model training set AUC was 0.745(95%CI: 0.649-0.844),the specificity and sensitivity were 0.811 and 0.603,and the verification set AUC was 0.771(95%CI:0.604-0.938),and the specificity and sensitivity were 0.667 and 0.935 respectively.The AUC of the joint model training set was 0.902(95%CI:0.846-0.957),the specificity and sensitivity were 0.865 and 0.781 respectively,and the AUC of the verification set was 0.871(95%CI:0.734-1.000),and the specificity and sensitivity were 0.800 and 0.935 respectively.The calibration curve shows that the predicted values of the joint model nomogram are in good agreement with the observed values.Decision curve analysis shows that nomogram has more clinical value than clinical model and imaging model.Conclusion: The texture analysis based on CT lung window image has a good value in predicting EGFR mutation of lung adenocarcinoma,and the nomogram combined with clinical and CT imaging features is beneficial to the realization of clinical accurate medical treatment.