Contrast-enhanced computed tomography-based intratumoral and peritumoral radiomics for identifying malignancy in pulmonary ground-glass nodules.

Background: Lung cancer remains the leading cause of cancer-related deaths worldwide. Early-stage lung cancer often presents as ground-glass nodules (GGNs) on computed tomography (CT). However, reliably distinguishing benign from malignant GGNs using conventional morphological features on CT remains a significant challenge, which impedes the accuracy of clinical surgical decision-making. Against this backdrop, radiomics, a technique that extracts quantitative features from medical images, offers a promising solution. This study aimed to investigate the predictive value of machine learning models based on radiomic features from intratumoral and peritumoral regions on contrast-enhanced CT in differentiating benign and malignant GGNs preoperatively. Methods: This retrospective study included 147 patients with pathologically confirmed GGNs who underwent contrast-enhanced CT before surgical resection between 2019 and 2023. Patients were randomly divided into a training set and a test set at a 7:3 ratio. Radiomics feature selection was performed using the minimum redundancy maximum relevance (mRMR) method followed by least absolute shrinkage and selection operator (LASSO) regression. Intratumoral and peritumoral models were built separately. Clinical features were selected via univariate and multivariate logistic regression analyses to construct a clinical model. An integrated radiomics-clinical model was then developed. Model performance was assessed using receiver operating characteristic (ROC) curves, sensitivity, specificity, calibration curves, and decision curve analysis (DCA). Results: Of the 147 patients, 87 had malignant and 60 had benign GGNs. The combined clinical-intratumoral model demonstrated the best diagnostic performance, with area under the ROC curve (AUC) values of 0.870 in both the training and test sets. The sensitivities were 0.787 and 0.654, and the specificities were 0.833 and 0.889, respectively. Conclusions: The intratumoral radiomics features based on enhanced CT, especially when combined with clinical features, have a high predictive value for the preoperative diagnosis of the benign and malignant nature of GGNs.