Abstract: Objective: To investigate the value of CT imaging radiomics features of different hepatic segments in predicting the severity of esophageal varices in cirrhotic patients. Methods: A total of 143 cirrhotic patients with portal hypertension and EV who were clinically confirmed and underwent gastroscopy from February 2018 to February 2021 were retrospectively selected. According to the severity of EV under gastroscopy, they were divided into severe group (61 cases) and non-severe group (82 cases). All patients received enhanced liver CT examination within 2 weeks before gastroscopy examination. The images of portal vein phase were selected and Shukun Radiomics V94 software was applied to extract the image radiomics features of caudate lobe, left lateral lobe, left inner lobe and right lobe (posterior upper segment + anterior upper segment) of liver at the level of bifurcations of portal vein in axial image according to Couinaud segmentation. Models to predict EV degree of different segments were established respectively. The value of radiomic features of different liver segments in predicting EV severity were compared. Results: Both severe and non-severe esophageal varices could be identified by the radiomics models established for predicting EV degree based on the CT imaging features of different hepatic segments in cirrhotic patients. The AUC values of the training set and verification set of the model of upper right lobe were:0.85 (0.78~0.92), 0.79 (0.64~0.93), respectively. That of caudate lobe, left lateral lobe, left inner lobe model training set and validation set were 0.78 (0.69~0.87) and 0.62 (0.44~0.79), 0.80 (0.71~0.88) and 0.64 (0.46~0.82), and 0.71 (0.61~0.81) and 0.73 (0.58~0.88), respectively. The AUC values of training set and validation set were the highest in the right lobe of liver model. Conclusion: All the CT images based radiomics features of different hepatic segments in cirrhotic patients could distinguish severe EV, among which the model of right hepatic lobe produces the best results.