Objective To explore the value of nomogram based on CT radiomics features and CT features for predicting Visceral Pleural Invasion(VPI)of invasive lung adenocarcinoma before operation.Methods Retrospective analysis of 234 patients with pathologically confirmed invasive lung adenocarcinoma.The patients were randomly divided into training group(n=164)and validation group(n=70)at 7∶3.The risk factors of VPI were screened by univariate and multivariate Logistic regression analysis in turn,and the CT model was constructed.The radiomics features of intratumoral(GTV),peri-tumor(PTV)and gross peritumoral tumor volume(GPTV)were extracted based on CT images,and the optimal feature subset and radiomics score were selected,and the optimal radiomics model was constructed and selected.The radiomics score of the best radiomics model combined with the CT features,the combined model was constructed and visualized by no-mogram.The effectiveness of each model was evaluated and compared with the receiver operating characteristic(ROC)curve and DeLong test.Decision curve(DCA)was used to evaluate the accuracy and clinical value of the model.Results Pleural thickening(P＜0.001)and tumor diameter(P＜0.01)were all CT risk factors for VPI.The AUC of the CT model is 0.74,0.81.The areas under the curve(AUC)of GPTV model for predicting VPI of IAC was 0.83,0.78,all higher than those of GTV model(AUC=0.78,0.70)and PTV model(AUC=0.81,0.74).AUC of combined model no-mogram(0.85,0.85)was higher than those of the CT model and GPTV model.Conclusion Nomogram based on CT fea-tures and GPTV radiomics features could effectively predict VPI.
NETL
NSTL
万方数据
