〔Abstract〕 Objective To explore the clinical dosimetry advantages of automated plan of IMRT for postoperative esophageal cancer and the dose prediction accuracy of the constructed 3D U-Res-Net model.Methods A total of110 postoperative esophageal cancer (middle and upper) cases treated by IMRT were considered in the study,of which 90 cases were randomly selected for training of deep learning prediction model.The deep learning prediction model and Auto-Plan module (Philips pinnacle316.2) were used to predict the three-dimension dose distribution and redesigned the remaining 20 cases respectively,and the results obtained were compared with manual plan.Results The average DSC value between the deep learning prediction plan and the manual plan was greater than0.92 in isodose surface,and the average Hausdorff distance HD95of the isodose surface was 0.58-0.62 cm;The V20,V30,Dmeanof total lung were slightly lower than those of manual plan (P<0.05) for the prediction model,meanwhile,the D2,D50,Dmean,HI of the target area and V30of total lungs were better than those of manual plan(P<0.05) for Auto-Plan;Three-dimensional dose distribution of the three groups and the corresponding DVH curve showed that the three-dimensional dose distribution of the three groups had a little differences,and the DVH curves of the target area and organs at risk had a good agreement.Conclusion Auto-Plan can realize the design of automated plan for postoperative esophageal cancer,while the deep learning prediction model can realize the accurate prediction of the 3D dose distribution.