Modification and establishment of CT criteria in preoperative assessment of portal venous invasion by hilar cholangiocarcinoma.

BACKGROUND: To compare the diagnostic performance of CT criteria and to establish a new model in evaluating portal venous invasion by hilar cholangiocarcinoma.

METHODS: CT images of 67 patients with hilar cholangiocarcinoma were retrospectively reviewed. Modified Loyer's, Lu's and Li's standard introduced from pancreatic cancer were used to evaluate portal venous invasion with the reference of intraoperative findings and/or postoperative pathological diagnosis. A new model was constructed with modified Lu's standard and contact length between portal vein and tumor.

RESULTS: The modified Loyer's standard, modified Lu's standard and Li's standard showed a sensitivity of 86.7%, 83.3%, 70.0%, a specificity of 89.4%, 95.7%, 95.7% and an accuracy of 88.6%, 92.0%, 88.1%, respectively. CT criteria performed better in evaluating left branch. The new model performed significantly better than any CT criterion or contact length, with a sensitivity of 95.0%, a specificity of 96.5% and an accuracy of 96.0%.

CONCLUSIONS: Modified Lu's standard performed best in evaluating portal venous invasion by hilar cholangiocarcinoma among three CT criteria. The left branch invasion could be evaluated by CT criteria better than the right branch and the trunk of portal vein. The new mode significantly improved the diagnostic performance of portal venous invasion by hilar cholangiocarcinoma.