Correlation Study Between Cavitary Lung Cancer And Cavitary Pulmonary Tuberculosis Based On Thin-section CT Radiomics Model

Objective:To explore the predictive value of thin-section CT-based Radiomics model for cavitary lung cancer and cavitary pulmonary tuberculosis.Methods:The clinical data and CT images of 92 patients(65 males,27 females)wi th pulmonary cavity were analyzed Retrospectively.These patients were treate d at multicenter represented by the First Hospital of Nanchang from February2017 to June 2021.All patients underwent routine thin-section chest CT scan s prior to therapy.These patients were randomly separated into the training set(74 cases)and the testing set(18 cases)by a 4:1 ratio.The 3D Slicer softwa re was used to extract 107 radiomics features of thin-section CT images,and then spearman correlation analysis and LASSO analysis were employed for fe ature selection.Logistic Regression method was used to construct the radiomi cs model,radiomics and clinical features model.Results:After delineating these lesions and feature extraction,five radiomics features were finally obtained from 107 features.In the radiomics model,the AUC values of the testing set and the training set were 0.75 and 0.86,respectively.Then,in the radiomics and clinical features model,the AUC values of the testing set and the training set were 0.93 and 0.90,respectively.The calibration curves showed the logistic model had good agreement between the predicted value and the actual value in the training set and testing set.And DCA curve demonstrated that the logistic model had good clinical applicability.Conclusion:The thin-section CT-based radiomics model has high diagnostic value in distinguishing cavitary lung cancer from cavitary pulmonary tuberculosis.