Fluorine-18 fluorodeoxyglucose PET/computed tomography-based radiomics for predicting risk groups, tumor stage, and capsular and fat tissue invasion in thymomas

PURPOSE: This study aimed to investigate the predictive value of fluorine-18 fluorodeoxyglucose PET/computed tomography (CT) based radiomics in patients with thymoma for risk stratification and tumor staging, including capsular and fat tissue invasion. METHODS: A total of 67 (50 for training and 17 for validation) patients with thymoma were analyzed. Thymomas were categorized as low risk (types A, AB, and B1) or high risk (types B2 and B3) and as early stage (I-II) or advanced stage (III-IV). PET- and CT-derived radiomic features extracted. Three models were constructed: a PET model, a CT model, and a PET/CT (combined) model. Feature selection was conducted using the Mann-Whitney U test and least absolute shrinkage and selection operator regression. RESULTS: For risk group discrimination, the combined model demonstrated the highest performance [area under the curve (AUC) = 0.80], outperforming the PET (AUC = 0.78) and CT (AUC = 0.72) models. In stage discrimination, the PET model and the combined model achieved superior performance (AUC = 0.82 for both), compared with the CT-based model (AUC = 0.75). For capsular invasion detection, predictive performance was achieved only with CT-based radiomics (AUC = 0.70). For fat tissue invasion detection, the combined model achieved the highest performance (AUC = 0.79), followed by CT (AUC = 0.76) and PET (AUC = 0.70). Maximum standardized uptake value of thymomas showed no significant discriminative value across any clinical categories (P > 0.05). CONCLUSION: F luorine-18 fluorodeoxyglucose PET/CT-based radiomics may provide complementary, noninvasive biomarkers for the management of patients with thymoma.