Prediction of Benign and Malignant Thymic Tumors based on Radiomics Features

The purpose of this study is to find out the radiomics features associated with differentiating benign and malignant thymic tumors from the quantitative CT features extracted by computer, and to establish a prediction mode, which can predict the benign and malignant thymic tumors, and improve the diagnostic efficiency of CT in thymic tumors. Retrospectively analyze the CT image data of 100 patients with pathologically confirmed thymic tumors. All selected images were preprocessed with A.K. (Artificial Intelligence Kit) software to extract the features of the lesions, and then we can use the Lasso algorithm to screen out CT radiomics features with diagnostic value and significant correlation with benign and malignant tumors. Logistic regression was used to construct a predictive model for the diagnosis of benign and malignant thymic tumors. The receiver operating characteristic(ROC) curve and area under curve (AUC) were used to evaluate the diagnostic efficacy of the constructed CT image features for benign and malignant thymic tumors. The most useful features of CT radiomics features were selected by Lasso algorithm: Surface Area, Inverse Difference Moment_AllDirection_offset1_SD, Voxel ValueSum, skewness, High Grey RuelnEmphasis_Allrection_offset7_SD. The area under ROC curve (AUC) was 0.752.The feature parameters screened on CT images have good diagnostic efficacy in differentiating benign from malignant thymoma.