Objective To establish models for predicting the benign and malignant pulmonary solid nodules based on plain CT imaging and plain & contrast-enhanced CT imaging,and to compare their efficacy.Methods A total of 854 pa-tients with pulmonary solid nodules in Tianjin Medical University Cancer Institute & Hospital from March 2013 to August 2017 were randomly divided into experimental group(n=598)and internal validation group(n=256)at a ratio of 7∶3.A total of 158 patients with pulmonary solid nodules in the Second Hospital of Shanxi Medical University from December 2018 to October 2019 were collected as an external validation group.Univariate and multivariate logistic regression were used to analyze the clinical and imaging characteristics,screen the predictive factors of the model,establish the plain CT based model and the plain & contrast-enhanced CT based model,and draw the nomograms.The receiver operating character-istic(ROC)curve and calibration curve of the model were drawn,and the area under ROC curve(AUC)of the models were calculated to evaluate the prediction efficiency of the models.Results The AUC of the plain CT based model and the plain & contrast-enhanced CT based model were 0.873(95％CI,0.843-0.898)and 0.912(95％CI,0.886-0.933)in the experimental group;and 0.875(95％CI,0.828-0.913)and 0.907(95％CI,0.864-0.940)in the internal valida-tion group;and 0.905(95％CI,0.849-0.946),0.926(95％CI,0.873-0.961)in the external validation group,re-spectively.Two models showed good prediction performance,and the plain & contrast-enhanced CT based model performed better.Conclusion The models based on clinical and CT imaging features have a high predictive performance,which can effectively predict the benign and malignant of pulmonary solid nodules before surgery.
维普
万方数据
