Construction and validation of a prediction model for malignant pulmonary nodules based on imaging, demographic, and laboratory features.

INTRODUCTION: This study aimed to explore the risk factors for malignant solitary pulmonary nodules (SPNs) and to develop an early integrated diagnostic model by analyzing patients' demographic data, clinical laboratory indicators, and imaging findings. METHODS: This retrospective study analyzed the medical records of 516 patients with a solitary pulmonary nodule(SPN) between December 2022 and August 2025. The cohort consisted of 402 patients with benign SPN and 114 with malignant SPN. Baseline characteristics were analyzed using SPSS 25 (version 25), and all subsequent statistical analyzes were performed using R 4.1.0. Variables were treated as categorical, with a two-sided p ≤ 0.05 considered statistically significant. Independent predictors of malignant SPN were identified using logistic regression and least absolute shrinkage and selection operator (LASSO) regression models. Model performance was evaluated and compared using the area under the receiver operating characteristic curve (ROC-AUC), calibration curves, and decision curve analysis (DCA). Based on a comprehensive comparison, the final predictive model was selected and presented as a nomogram to facilitate individualized clinical decision-making. RESULTS: Logistic regression analysis identified five independent predictors for the benign or malignant status of pulmonary nodules:age (p = 0.04), smoking history (p = 0.02), nodule diameter (p = 0.01), lobulation (p = 0.02), and spiculation (p = 0.02). The least absolute shrinkage and selection operator (LASSO) regression model indicated that spiculation, nodule diameter, carcinoembryonic antigen, and smoking history possessed independent predictive value. Based on a comparison of predictive performance, the logistic regression model (both univariate and multivariate) demonstrated superior discrimination and calibration compared to the LASSO regression model. Consequently, a visual nomogram was developed based on the results of the logistic regression analysis. CONCLUSION: This study employed dual modeling with logistic regression and LASSO regression. The analysis identified five independent predictors for malignant SPN: age, smoking history, nodule diameter, lobulation, and spiculation. In comparison, the logistic regression model demonstrated superior discrimination and calibration. Based on these findings, a visual nomogram was developed to facilitate clinical application.