Objective To investigate the clinical value of CT-based extracellular volume fraction (ECV) combining with CT features for the preoperative pathological classification of intrahepatic cholangiocarcinoma (ICC; Large duct-ICC, LD-ICC；Small duct-ICC, SD-ICC).Methods From March 2018 to January 2024, sixty-four patients with pathologically confirmed ICC who had available preoperative abdominal plain and enhanced CT images were retrospectively enrolled.The ECV of hyperenhancement region (ECV1) and that of hypoenhancement region (ECV2) in tumor was calculated and the relative ECV (ECVr) was defined as their difference. ECVr, CT features as well as clinical data, for differentiating LD-ICC from SD-ICC were compared using univariate and multivariate analyses, and a logistic regression model was then generated. The differential diagnostic efficiency of the logistic regression model was further evaluated. Results The univariate analysis showed that higher ECVr, fewer complete contours in arterial phase hyperenhancement, central type, intrahepatic duct dilatation were suggestive factors of LD-ICC. Higher ECVr and intrahepatic duct dilatation, tumor location were independent factors for LD-ICC in multivariate analysis. The logistic regression model has achieved a favorable preoperative differential diagnositic performance in ICC pathological categorization. The AUC of the logistic regression model was up to 0.928 (95% CI: 0.866, 0.991), with the sensitivity of 84.0% and the specificity of 92.3%. Conclusion Higher ECVr, intrahepatic duct dilatation, tumor location can be considered reliable CT indicators of LD-ICC. CT combined model can facilitate noninvasive prediction of ICC subtype with satisfactory predictive performance