Analysis of MRI features correlated with TNM staging in thymic epithelial tumors

This study aimed to investigate the association between conventional MRI features and TNM staging in thymic epithelial tumors (TETs), as well as to evaluate their ability to differentiate early-stage from advanced-stage disease. Clinical data and MRI features of 69 patients with pathologically confirmed TETs were retrospectively analyzed. Patients were categorized into early-stage (stage Ⅰ-Ⅱ) and advanced-stage (stage Ⅲ-Ⅳ) groups according to the 9th edition TNM staging system. Between-group comparisons were used to evaluate differences in MRI features between the two groups. Multivariable logistic regression was performed to identify MRI features associated with advanced-stage disease, and receiver operating characteristic (ROC) curve was performed to assess diagnostic performance. Progression-free survival (PFS)-related factors were explored using univariable Cox regression analysis. Advanced-stage TETs showed a larger maximum tumor diameter and more frequent aggressive MRI features, including pleural effusion, heterogeneous T2WI signal intensity, restricted diffusion, heterogeneous enhancement, internal septation, fat space invasion, pleural/pericardial invasion, and major vascular invasion, than early-stage TETs (all P < 0.05). In multivariable logistic regression model, fat space invasion and major vascular invasion were independently associated with advanced-stage disease (P < 0.05). The combined model based on these two features achieved an area under the curve (AUC) of 0.895, with a sensitivity of 85.7% and specificity of 90.0% for differentiating advanced-stage from early-stage TETs. Conventional MRI features, particularly fat space invasion and major vascular invasion, were associated with advanced TNM stage in TETs and may provide supportive information for preoperative staging assessment. Given the limited sample size and exploratory nature of the survival analysis, these findings require validation in larger multicenter cohorts.