Differentiation of AFP-negative hepatocellular carcinoma from other intrahepatic malignant lesions by a noninvasive predictive model based on Sonazoid contrast-enhanced ultrasound.

Objectives: This study aimed to develop and validate a non-invasive predictive model, which was a reliable nomogram to accurately differentiate AFPN-HCC from other intrahepatic malignant lesions. Methods: This study enrolled 165 patients with malignant focal liver lesions, including AFPN-HCC (n=85) and other intrahepatic malignant lesions (n=80). Data were analyzed to screen for risk factors phase by using LASSO regression as well as univariate and multivariate logistic regression analysis. We constructed a model and developed a nomogram. Then using the area under the curve, Hosmer-Lemeshow test, calibration curves, decision curve analysis, and 1,000 bootstraps to assess and internally validate the model performance. We calculated the optimal threshold, sensitivity, specificity, positive and negative predictive value, and accuracy of the prediction model. Results: LASSO and multivariate logistic regression analyses indicated that tumor number, necrosis in tumor, arterial phase enhancement pattern, arterial phase perfusion velocity, and Kupffer phase degree of washout were the significant predictors to differentiate AFPN-HCC from OM. The AUC was 0.886, and the AUC of internal validation was 0.865. The optimal critical value of the predicted value was 0.524, with a sensitivity of 82.35%, specificity of 85.00%, positive predicted value of 85.37%, negative predicted value of 81.93%, and an accuracy of 83.64%. The Conclusion: Our nomogram has excellent discrimination, calibration and clinical utility by combining SCEUS and clinical features, which may help clinicians improve the diagnostic performance for AFPN-HCC, contributing to individualized treatment.