Based On CT Features Nomogram For The Prediction Of Invasive Pulmonary Adenocarci Nomas In Pure Ground-glass Nodule

Objective:Retrospectively investigated the pulmonary pure ground-glass nodule(pGGN),to identify the risk factors and construct a nomogram for distinguishing preinvasive lesion with invasive pulmonary adenocarcinomas(IPA),and the calibration curves for the probability were drawn in the development and validation cohorts.Methods:A primary cohort of patients with surgery pathologically confirmed pulmonary pure ground-glass nodlue were retrospectively studied at two local institutions from January 2015 to June 2018.Two thirds(n=180)were randomly selected and assigned to a development group and the remaining patients(n=120)were assigned to a validation group.Record and analyze the general clinical information,pathological types(preinvasive lesion or invasive pulmonary adenocarcinomas),various CT features.Statistical differences between preinvasive lesions and IPA were analyzed by using the independent sample t test for finding out the differences in patient's age,lesion size,and mean CT value.Statistical differences between patient's sex and CT features(eg:lesion shape,margin,abnormal air bronchogram,the relationship between blood vessels and pure GGNs,and bubble lucency)were analyzed by using the Pearson?~2 test and Fisher exact test,as appropriate.Characteristics with a P value of less than 0.05 at univariate analysis were used as input variable for binary logistic regression analysis.A nomogram constructed by R3.5.1 with the rms statistical packages.A nomogram to predict IPA was developed,and the calibration curves for the probability were drawn.The ROC analysis revealed that the optimal cutoff of Risk value identified IPA from preinvasive lesion.Results:1)Compared to characteristic of patients:in development group,there were significant differences in lesion size,mean CT attenuation value,shape,margin,abnormal air bronchogram,the relationship between blood vessels and pure GGNs,and bubble lucency(P=0.000,0.028,0.000,0.000,0.000,0.000,0.029,separately).No difference were found in sex(P=0.107),ages(P=0.412),pleural indentation(P=0.359).2)Independent risk factors:after binary logistic regression analysis,the size of the lesion(OR=76.69,95%CI:11.540-513.286,P=0.000)and mean CT attenuation value(OR=1.006,95%CI:1.002-1.011,P=0.005)were risk factors and entered into the nomogram.3)Performance of nomogram:the nomogram showed good discrimination for internal validation with C-index of 0.841[95%CI:0.785-0.898]and for external validation with C-index of 0.836[95%CI:0.766-0.907],and had well-fitted calibration curves.In the development cohort,the optimal cutoff risk value was 0.5325 for predicting IPA,and the sensitivity was 81.82%,specificity was 71.43%.In the validation cohort,the optimal cutoff risk value was 0.5404 for predicting IPA,and the sensitivity was65.15%,specificity was 88.89%.Conclusion:1)For pure GGN,lesion size,mean CT attenuation value,shape,margin,internal air bronchogram,the relationship between blood vesselsthe can be discriminators of preinvasive lesions from IPAs.2)For pure GGN,lesion size and mean CT value were the risk factors in differentiating preinvasive lesion from IPA.3)Based on lesion size and mean CT value,a nomogram that can predict the risk of IPA for patients with pure GGN.