PURPOSE: To compare the accuracy of computed tomography (CT) with that of digital subtraction angiography (DSA) in predicting the resectability of Klatskin tumor on the basis of vascular invasion. MATERIALS AND METHODS: Twenty-five patients with Klatskin tumor who had undergone laparotomy were in-cluded in this study. In order to assess the surgical resectability of their tumors, the preoperative CT scans and DSA of these patients were retrospectively assessed in terms of vascular invasion. The criteria of unresectability were tumoral invasion of the proper hepatic artery or main portal vein, or simultaneous invasion of the hepatic artery on one side and the other side portal vein. RESULTS: Tumors were unresectable in 13 cases, and resectable in 12. CT and DSA predicted nine and three tumors as unresectable ones, respectively. The sensitivity, specificity, positive predictive value, negative predic-tive value and accuracy of CT in determining whether a tumor was unresectable were 61.5 %, 91.7%, 88.9 %, 6 8 .8 % and 76.0 %, respectively. For DSA, the respective figures were 23.1 %, 100 %, 100 %, 54.5 % and 6 0 .0 %. For the detection of vascular invasion without diameter change, CT was superior to DSA; for the evaluation of vascular anatomy, it was, however, less effective. CT failed to detect small hepatic metastasis (n=2), lymph node metastasis (n=1), variation of the bile duct (n=1), and the distal extent of tumor in the bile duct (n=1), factors which precluded surgical resection. CONCLUSION: CT is a reliable method for the detection of vascular invasion and tumor unresectability. For the detection of vascular anatomic variation, the combined use of CT and DSA would be helpful.