Abstract: Objective: To construct and validate computed tomography (CT) radiomics models to preoperatively predict regional lymph node (LN) status of resectable adenocarcinoma of the esophagogastric junction (AEG). Methods: In total, 270 consecutive patients with AEG were enrolled in this study. Of these, 220 patients from Institution A were stratified into training (n=153) and test cohorts (n=67). The remaining 50 patients from Institution B were assigned to the external validation cohort. Within the training cohort, preoperative CT radiomics features extracted from the AEGs and LNs were used to construct the AEG and LN radiomics models, respectively; the radiomics scores (RS) of the AEGs and LNs were integrated with the clinical features to build the combined model. The predictive performances of the individual models were evaluated using the area under the receiver operating characteristic (AUROC) curve. The DeLong test was used to compare the predictive performance of the models. Results: Ten AEG and four LN radiomics features were screened to develop the AEG and LN radiomics models for predicting LN status, respectively. The combined model was developed by integrating AEG-RS and LN-RS with cT-stage and it achieved higher AUROC curve values than the AEG or LN radiomics models, alone, for the training (0.925 vs. 0.755 or 0.857), test (0.897 vs. 0.716 or 0.836), and external validation (0.935 vs. 0.706 or 0.849) cohorts. The DeLong test showed that the predictive performance of the combined model was significantly superior to that of the AEG and LN radiomics models, alone, in the three cohorts (all P <0. 05), and the predictive performance of the LN radiomics model was significantly superior to that of the AEG radiomics model in the three cohorts (all P < 0.05). Conclusion: Based on the radiomics method, the combined model is effective at preoperatively evaluating the regional lymph node status of patients with AEG.