Analysis of Sonazoid contrast-enhanced ultrasound for predicting the risk of microvascular invasion in hepatocellular carcinoma: a prospective multicenter study.

Abstract

The aim of this study was to evaluate the potential of Sonazoid contrast-enhanced ultrasound (SNZ-CEUS) as an imaging biomarker for preoperative prediction of microvascular invasion (MVI) in hepatocellular carcinoma (HCC). From August 2020 to March 2021, we conducted a prospective multicenter study on the clinical application of Sonazoid in liver tumor; a MVI prediction model was developed and validated by integrating clinical and imaging variables. Multivariate logistic regression analysis was used to establish the MVI prediction model; three models were developed: a clinical model, a SNZ-CEUS model, and a combined model and conduct external validation. We conducted subgroup analysis to investigate the performance of the SNZ-CEUS model in non-invasive prediction of MVI. Overall, 211 patients were evaluated. All patients were split into derivation (n = 170) and external validation (n = 41) cohorts. Patients who had MVI accounted for 89 of 211 (42.2%) patients. Multivariate analysis revealed that tumor size (> 49.2mm), pathology differentiation, arterial phase heterogeneous enhancement pattern, non-single nodular gross morphology, washout time (< 90s), and gray value ratio (≤ 0.50) were significantly associated with MVI. Combining these factors, the area under the receiver operating characteristic (AUROC) of the combined model in the derivation and external validation cohorts was 0.859 (95% confidence interval (CI): 0.803-0.914) and 0.812 (95% CI: 0.691-0.915), respectively. In subgroup analysis, the AUROC of the SNZ-CEUS model in diameter  ≤ 30mm and ˃30mm cohorts were 0.819 (95% CI: 0.698-0.941) and 0.747 (95% CI: 0.670-0.824). Our model predicted the risk of MVI in HCC patients with high accuracy preoperatively. Sonazoid, a novel second-generation ultrasound contrast agent, can accumulate in the endothelial network and form a unique Kupffer phase in liver imaging. The preoperative non-invasive prediction model based on Sonazoid for MVI is helpful for clinicians to make individualized treatment decisions. • This is the first prospective multicenter study to analyze the possibility of SNZ-CEUS preoperatively predicting MVI. • The model established by combining SNZ-CEUS image features and clinical features has high predictive performance in both derivation cohort and external validation cohort. • The findings can help clinicians predict MVI in HCC patients before surgery and provide a basis for optimizing surgical management and monitoring strategies for HCC patients.