Abstract:Objective To investigate the predictive value of preoperative CT signs and imaging parameters for the spread through air spaces（STAS）in pT1a-pT1c stage lung adenocarcinoma，and to provide a reference for designing reasonable surgical plans for lung cancer patients. Methods Patients with pT1a-pT1c stage lung adenocarcinoma who were admitted to Shijiazhuang People's Hospital from December 2021 to December 2024 were selected. They were divided into STAS-positive and STAS-negative groups based on postoperative pathological diagnosis results. The clinical data，preoperative CT signs，and imaging parameters of the two groups were compared，and the relationship of preoperative CT signs and imaging parameters with STAS was analyzed. An XGBoost model was constructed based on preoperative CT signs and imaging parameters. The predictive value of the XGBoost model was evaluated using calibration curves，receiver operating characteristic（ROC）curves，and decision curve analysis（DCA）；additionally，103 patients with pT1a- pT1c stage lung adenocarcinoma from the same period were selected as the external validation set for external validation. Results Ultimately，237 patients with lung adenocarcinoma were included as study subjects，including 55 cases in the STAS positive group and 182 cases in the STAS negative group. A significant difference was observed in the clinical T-stage between the two groups（P＜0.05）. In the STAS-positive group，the incidence of lobulated sign，spiculated sign，vascular convergence sign，bronchial air bronchogram，and tumor diameter，solid component diameter，chord-arc distance/chord length ratio，and net increase in arterial phase were higher than those in the STAS-negative group（P＜0.05），and the above CT signs and imaging parameters were positively correlated with STAS（P＜0.05）. The XGBoost algorithm identified six factors to construct a predictive model，ranked in order of importance as follows：chord-arc distance/chord length ratio，solid component diameter，vascular convergence sign，net increase in arterial phase，spiculated sign，and bronchial air bronchogram. The calibration curve indicated that the calibration degree of the XGBoost model for preoperative CT signs and imaging parameters was 0.871，with a consistency index of 0.884，suggesting high predictive accuracy of the model. The ROC curve showed that the area under the curve（AUC）of the XGBoost model for predicting STAS was 0.899（95%CI：0.860-0.937），indicating good predictive performance of the model. The DCA curve demonstrated that when the threshold probability range was between 10% and 100%，the XGBoost model for predicting STAS had a high net benefit. The sensitivity，specificity，and Kappa value of the XGBoost model for predicting STAS in the external validation set were 87.50%，93.67%，and 0.789（95%CI：0.596-0.982），respectively，indicating high consistency with clinical actual results（P＜0.05）. Conclusion The XGBoost model，constructed based on preoperative CT features including the diameter of solid components，vascular convergence sign，chord-arc distance/chord length ratio，net increase in arterial phase，bronchial air bronchogram，and spiculation sign，exhibits high application value for predicting STAS in pT1a-pT1c stage lung adenocarcinoma，and can provide guidance information for clinical surgical planning.