Objective:To construct a predictive model based on the clinical and imaging features of lung adenocarcinoma to predict occult mediastinal lymph node metastasis. Methods:A retrospective analysis was conducted on the patients of First Affiliated Hospital of Nanjing Medical University from 2009 to 2019 who underwent surgical treatment and lymph node dissection with/without occult mediastinal lymph node metastasis on pathology.Multiple clinical and imaging features of the patient were collected.Univariate and multivariate analyses were used to screen independent predictors and construct imaging models for multiple CT characteristics.A receiver operating characteristic (ROC) was established to evaluate the predictive efficacy and clinical utility value of each model. Results:Of the 2,287 patients , there were 780 cases of lung adenocarcinoma, of which 145 had lymph node metastases.Univariate analysis suggested that tumor size, axial location, nodal nature, morphological features, pleural pulling sign, and type of pleural adjacency were significantly associated with lymph node metastasis.Multivariate analysis indicated that tumor size (OR = 1.019; 95% CI: 1.002-1.036; p = .028), nodal nature (OR = 0.361; 95% CI: 0.202-0.646; p = .001), pleural involvement (OR = 1.835; 95% CI: 1.152-2.924; p = .011), and presence of mediastinal pleural adjacency (OR = 1.796; 95% CI: 1.106-2.919; p = .018) were independent predictors of occult mediastinal lymph node metastasis.The AUC value of the imaging model was 0.75, with the sensitivity and specificity of 86.2% and 53.1%, respectively. Conclusion:The imaging model based on chest CT plain scan shows an excellent clinical value in predicting occult mediastinal lymph node metastasis of lung adenocarcinoma，which can provide clinicians with a basis for noninvasive preoperative decision making and help thoracic surgeons choose more appropriate surgical treatment options.