LEI Chong 1 , XU Ying 2 , CHEN Wei 1 , DU Jin 1 , SUN Jie 1 , LI Xin 1 , LIU Haixia 1
  • 1. Department of CT, Tangshan Gongren Hospital, Tangshan, 063000, Hebei, P. R. China;
  • 2. Department of Functional Examination, Tangshan Gongren Hospital, Tangshan, 063000, Hebei, P. R. China;
LIU Haixia, Email: haixia_liu69@163.com
CT; blood cell parameter; tumor marker; pure ground-glass nodule; infiltration
10.7507/1007-4848.202407062
Export PDF Favorites Scan Get Citation

Abstract Full text Figures/Tables References Cited by

Objective  To develop and assess the performance of a predictive model for the infiltration degree of solitary pulmonary pure ground-glass nodules (pGGN) using CT, blood cell parameters, and tumor markers. Methods  The clinical data of patients with solitary pulmonary pGGN, collected from Tangshan Gongren Hospital between June 2021 and April 2024, were analyzed. They were divided into a training set and a test set in a 7∶3 ratio. Lasso-logistic regression was used to identify risk factors for invasive adenocarcinoma and construct the model. The model's performance was assessed using receiver operating characteristic (ROC) curves, calibration curves, mean absolute error (MAE), mean squared error (MSE), and accuracy. Results  The study included 528 patients (265 males, 263 females) with a median age of 54 years (interquartile range: 45-59). Lasso-logistic regression identified increased diameter, vascular convergence sign, pleural indentation sign, elevated mean CT value, and elevated carcinoembryonic antigen levels as independent risk factors for solitary pulmonary pGGN infiltration. In contrast, a rounded or similarly rounded shape and an elevated platelet-to-lymphocyte ratio were independent protective factors (P<0.05). In the training set, the area under the ROC curve of model Z (comprising diameter, vascular convergence sign, pleural indentation sign, rounded or similarly rounded, mean CT value, carcinoembryonic antigen, and platelet-to-lymphocyte ratio) was 0.875, which was greater than that of model C (comprising diameter, vascular convergence sign, pleural indentation sign, rounded or similarly rounded, and mean CT value; 0.852) and model S (comprising carcinoembryonic antigen and platelet-to-lymphocyte ratio; 0.753). The MAE, MSE, and accuracy of model Z were 0.035, 0.003, and 0.808, respectively, which were lower than those of model C (0.058, 0.006, and 0.827) and higher than those of model S (0.031, 0.001, and 0.648). In the test set, the area under the ROC curve, MAE, MSE, and accuracy of model Z were 0.829, 0.051, 0.004, and 0.755, respectively, which were higher than those of model C (0.780, 0.038, 0.002, and 0.730) and model S (0.740, 0.042, 0.002, and 0.692). Conclusion  The model constructed from diameter, vascular convergence sign, pleural indentation sign, rounded or similarly rounded shapes, mean CT value, carcinoembryonic antigen, and platelet-to-lymphocyte ratio aids in assessing the infiltration degree of pulmonary pGGN, with superior performance compared to models based solely on CT or those based on tumor markers combined with blood cell parameters.