Development of a CT-based intra- and peritumoral clinical-radiomics nomogram for differentiating lung squamous cell carcinoma and adenocarcinoma

This study aimed to develop and validate a CT-based intra- and peritumoral clinical-radiomics nomogram for preoperative differentiation of lung squamous cell carcinoma (LUSC) from lung adenocarcinoma (LUAD). A total of 278 histopathologically confirmed lung cancer patients (157 LUSC, 121 LUAD) were retrospectively enrolled and randomly divided into a training set of 195 cases and a testing set of 83 cases. Radiomic features were extracted from intratumoral regions and peritumoral regions expanded by 3 mm and 5 mm. Feature selection was performed using the Mann-Whitney U test and LASSO regression. Four machine learning classifiers were compared, and logistic regression was selected as the optimal classifier. A combined clinical-radiomics nomogram was then constructed using radiomics score, age, and tumor size, and model performance was assessed by AUC, calibration curves, and decision curve analysis (DCA). The logistic regression-based intratumoral radiomics model achieved AUCs of 0.946 in the training set and 0.858 in the testing set. Peritumoral expansion (3 mm and 5 mm) did not significantly improve performance. The combined clinical-radiomics model based on intratumoral features showed the best performance, with AUCs of 0.969 (training) and 0.911 (testing), and the nomogram demonstrated good calibration and clinical utility. In conclusion, the proposed CT-based intratumoral clinical-radiomics nomogram can effectively distinguish LUSC from LUAD, offering a non-invasive and valuable tool to guide personalized treatment decisions for lung cancer patients.