Objectives: The aim of this study was to develop a predictive model based on Sonazoid contrast-enhanced ultrasound (SCEUS) and clinical features to discriminate poorly differentiated hepatocellular carcinoma (P-HCC) from intrahepatic cholangiocarcinoma (ICC).
Patients and method: Forty-one ICC and forty-nine P-HCC patients were enrolled in this study. The CEUS LI-RADS category was assigned according to CEUS LI-RADS version 2017. Based on SCEUS and clinical features, a predicated model was established. Multivariate logistic regression analysis and LASSO logistic regression were used to identify the most valuable features, 400 times repeated 3-fold cross-validation was performed on the nomogram model and the model performance was determined by its discrimination, calibration, and clinical usefulness.
Results: Multivariate logistic regression and LASSO logistic regression indicated that age (> 51 y), viral hepatitis (No), AFP level (≤ 20 µg/L), washout time (≤ 45 s), and enhancement level in the Kupffer phase (Defect) were valuable predictors related to ICC. The area under the receiver operating characteristic (AUC) of the nomogram was 0.930 (95% CI: 0.856-0.973), much higher than the subjective assessment by the sonographers and CEUS LI-RADS categories. The calibration curve showed that the predicted incidence was more consistent with the actual incidence of ICC, and 400 times repeated 3-fold cross-validation revealed good discrimination with a mean AUC of 0.851. Decision curve analysis showed that the nomogram could increase the net benefit for patients.
Conclusions: The nomogram based on SCEUS and clinical features can effectively differentiate P-HCC from ICC.
Keywords: Intrahepatic cholangiocarcinoma; Nomogram; Poorly differentiated hepatocellular carcinoma; Sonazoid contrast-enhanced ultrasound.
© 2023. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.