Objectives To explore the application value of diffusion-weighted imaging (DWI) based radiomics column chart (RN) in predicting the beneficiaries of adjuvant chemotherapy (AC) after partial hepatectomy (HPx) for intrahepatic cholangiocarcinoma (ICC). Methods A retrospective analysis was conducted on the MRI and clinical pathological data of 85 ICC patients who underwent HPx in our hospital from January 2018 to January 2022, divided into a training set of 60 cases and a validation set of 25 cases. Using the random forest algorithm to screen imaging omics features based on DWI and establish an imaging omics model (RS) for predicting early recurrence (ER), and calculating the imaging omics score. Multivariate logistic regression method for screening clinical, pathological, and conventional imaging factors establish a clinical pathological imaging (CPR) model for ER related variables. Construct a clinical radiomics column chart (RN) based on the selected variables and radiomics scores, evaluate and verify the discrimination and calibration of the above model. Use Kaplan Meier method and log rank test to analyze the differences in prognosis among different groups. Results In the validation set, the DMI based RS and CPR models have similar discriminative performance in predicting ER (AUC: 0.762 vs 0.624, P=0.271), while the RN model (AUC: 0.833) has significantly better predictive performance than RS (P=0.012). In the high-risk group of ER patients predicted by the RN model, OS and DFS of patients who underwent surgery+AC were significantly longer than those who only received surgical treatment (DFS, P=0.035; OS, P=0.020). Conclusion The RN model based on DWI can effectively predict the postoperative occurrence of ICC patients ER and also effectively screen the beneficiaries of postoperative AC