Contrast-enhanced CT-based radiomics differentiate anterior mediastinum lymphoma from thymoma without myasthenia gravis and calcification

To explore the value of a radiomics model based on enhanced computed tomography (CT) in differentiating anterior mediastinal lymphoma (AML) and thymoma without myasthenia gravis (MG) and calcification. The present study analysed patients who were diagnosed histologically with AML and thymoma in three independent institutions. All pre-treatment patients underwent enhanced CT. In the training group of patients from institutions 1 (the First Affiliated Hospital of Kunming Medical University) and 3 (the Yunnan Cancer Hospital), two radiologists independently analysed the enhanced CT images and performed manual segmentation of each tumour. Radiomics features were screened using interobserver interclass coefficient (ICC) analysis, feature correlation analysis, and L1 regularisation. The discriminative efficacy of the logistic regression model was evaluated using receiver operating characteristic (ROC) analysis. Validation group of patients from institution 2 (the Second Affiliated Hospital of Zhejiang University School of Medicine) was used to validate the proposed models. A total of 114 patients were enrolled in this study and 1,743 radiomics features were extracted from the enhanced CT images. After feature screening, the remaining 37 robust radiomics features were used to construct the model. In the training group, the AUC of the model was 0.987 (95% confidence interval [CI]: 0.976–0.999), the sensitivity, specificity, and accuracy were 0.912, 0.946, and 0.924, respectively. In the validation group, the AUC of the model was 0.798 (95% CI: 0.683–0.913), the sensitivity, specificity, and accuracy were 0.760, 0.700, and 0.743, respectively. The radiomics model created provided effective information to assist in the selection of clinical strategies, thus reducing unnecessary procedures in patients with AML and guiding direct surgery in patients with thymoma to avoid biopsy. •Multicenter acquisition of lymphoma and thymoma cases.•Explore the effective radiomics features for distinguishing lymphoma from thymoma.•Myasthenia gravis and calcification were firstly proposed as exclusion criteria.