Objective To explore the value of CT radiomics combined with different machine learning algorithms in preoperative pre-diction of spread through air spaces(STAS)status of stage Ⅰ lung adenocarcinoma.Methods A total of 183 patients with stage Ⅰ primary lung adenocarcinoma were retrospectively selected,83 in the STAS-positive group and 100 in the STAS-negative group.The clinical data and CT features were compared between the two groups.All patients were randomly divided into training set(n=146)and val-idation set(n=37)at the ratio of 8∶2.ITK-SNAP software was used to segment the region of interest(ROI).Radiomics features were extracted by PyRadiomics.Three-step feature selection was used for feature dimensionality reduction and six machine learning algorithms were used to construct the prediction models.The receiver operating characteristic(ROC)curve was used to assess the predictive efficacy of the 6 models.Decision curve analysis(DC A)was used to evaluate the clinical application value of the models.Results There were significant differences in tumor diameter,density,presence of micropapillary/solid components,vascular inva-sion,pleural invasion,and Ki-67 expression between the two groups(P＜0.05).Multivariate logistic regression analysis showed tumor density,presence of micropapillary/solid components,and vascular invasion were independent predictors of STAS.The area under the curve(AUC)of the multilayer perceptron(MLP)model were 0.806 and 0.841 in the training set and validation set,respectively.DCA showed good clinical decision benefits from the model.Conclusion The combined model based on CT radiomics and MLP can effectively predict STAS of stage Ⅰ lung adenocarcinoma before operation.It can provide decision support for diagnosing and treating early-stage lung adenocarcinoma.