Super‐resolution PET/CT radiomics nomogram for predicting spread through air spaces in stage I lung adenocarcinoma

BACKGROUND: Super-resolution (SR) reconstruction-based positron emission tomography (PET) imaging has been widely applied in the field of computer vision. However, their definitive clinical benefits have yet to be validated. Radiomics-based modeling provides an effective approach to evaluate the clinical utility of SRPET imaging. PURPOSE: F]FDG PET/CT) in predicting the status of spread through air spaces (STAS) preoperatively in patients with clinical stage I lung adenocarcinoma (LUAD). METHODS: A total of 131 clinical stage I lung cancer patients were retrospectively included and randomly divided into two cohorts: training (n = 91) and test (n = 40). A transfer learning network enhanced PET image resolution to produce preoperative SRPET images. Radiomics features were extracted from SRPET, PET, and CT images. A radiomics nomogram was developed using clinically independent predictors and the optimal radiomics signature. Its predictive performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). RESULTS: Five models were constructed to predict STAS status. Among these, the comprehensive model-which integrated 1 clinical feature, 6 CT features, and 14 SRPET features-demonstrated the highest area under the curve (AUC) values of 0.948 in the training cohort and 0.898 in the test cohort. It outperformed previous models in net benefits on calibration and decision curves. These findings support developing a nomogram for visualizing STAS prediction preoperatively. CONCLUSION: The SRPET/CT radiomics nomogram effectively predicted STAS in clinical stage I LUAD and may aid in guiding individualized therapy plans before surgical intervention.