Objectives: To assess the usefulness of classification tree analysis (CTA) for discrimination of hepatocellular carcinoma (HCC) with target sign on hepatobiliary phase (HBP) and/or diffusion-weighted image (DWI) from intrahepatic cholangiocarcinoma (ICC).
Methods: This retrospective study was approved by the institutional review board. From among the 811 patients with histopathologically proven 79 ICCs and 732 HCCs, 69 patients with 69 (87.3%) ICCs and 115 patients with 115 HCCs (15.7%) including 25 scirrhous HCCs and 10 HCCs with central scar that showed the target sign either on HBP or on DWI were included. Two radiologists evaluated the presence of capsule, septum, and arterial enhancement pattern on MR imaging. Capsule, septum, arterial enhancement pattern, and target sign on HBP or DWI were used to classify the tumors using CTA.
Results: On CTA, capsule was the initial predictor for assessing the probability of tumors being HCC. The CTA model demonstrated a sensitivity of 86.1%, specificity of 76.8%, and accuracy of 82.6% for discriminating between ICCs and HCCs. In 115 HCCs, only 16 (13.9%) tumors were misclassified as high probability of ICC, and 64.0% (16/25) scirrhous HCCs and 90.0% (9/10) HCCs with central scar were correctly classified as high probability of HCC.
Conclusions: Target sign either on HBP or on DWI was shown in 87.3% (69/79) of ICCs and 15.7% (115/732) of HCCs. The CTA applying capsule and septum may be useful for guiding correct diagnosis of atypical HCCs with the target sign from ICCs.
Keywords: Cholangiocarcinoma; Diffusion-weighted image; Gadoxetic acid; Hepatobiliary phase; Hepatocellular carcinoma; Target sign.
Copyright © 2017. Published by Elsevier B.V.