Assessment of pulmonary nodule characteristics by using Lung RADS,C Lung RADS and 3D reconstruction features

DONG Li; SHENG Mao; TAO Lei; CHEN Lijie; TANG Ming; HE Jiawei

doi:10.12122/j.issn.1674-4500.2025.07.11

Abstract：Objective To compare the diagnostic value of Lung RADS and C Lung RADS in characterizing pulmonary nodules,and to evaluate the predictive performance of C Lung RADS in combination with 3D reconstruction features.Methods A retrospective analysis was conducted on clinical and 3D imaging data from 153 patients who underwent lobectomy for pulmonary nodules at our institution from January 2023 to January 2025.Independent predictive factors were identified using least absolute shrinkage and selection operator(LASSO)regression and multivariate logistic regression.The diagnostic performance of Lung RADS and C Lung RADS was compared using Venn diagrams and confusion matrices.A combined predictive model was constructed and evaluated using ROC curves,calibration curves,Hosmer-Lemeshow goodness-of-fit tests,decision curve analysis,and clinical impact curves.Results Both systems showed high concordance in the high-risk and very-high-risk categories.Three-dimensional reconstruction features significantly differed between benign and malignant nodules(P<0.05),with average CT value and intranodular vascularity emerging as independent predictors.The diagnostic performance of three-dimensional features combined with either Lung RADS or C Lung RADS did not differ significantly(PDelong=0.341),while the combined model significantly outperformed the three-dimensional features alone(PDelong=0.020).Ultimately,the model integrating three-dimensional features with C Lung RADS was selected,yielding an AUC of 0.875(95%CI:0.786-0.963),a goodness-of-fit test(χ²=9.825,P=0.278),and demonstrated clinical net benefit within a risk threshold range of 0.02-0.58 based on decision curve analysis and clinical impact curves.Conclusion Lung RADS and C Lung RADS performed similarly in assessing pulmonary nodule malignancy risk,but C Lung RADS offered a more concise classification scheme.The combined model incorporating 3D reconstruction features and C Lung RADS demonstrated favorable predictive performance and holds promise for aiding in the preoperative differentiation of benign and malignant pulmonary nodules.