关键词: 小肝癌, 微血管侵犯, 灰阶超声, 超声造影, 预测模型

Abstract: Objective To observe the gray-scale and contrast-enhanced ultrasound (CEUS) features of microvascular invasion (MVI) in small hepatocellular carcinoma (sHCC) under the background of liver cirrhosis, and to analyze the risk factors for MVI in these patients, and construct a prediction model. Methods A retrospective analysis was conducted on the data of 120 patients with sHCC confirmed by surgery and pathology collected from June 2020 to June 2024. The patients were divided into a MVI group and a non-MVI group according to whether they had MVI. Gray-scale ultrasound and CEUS features of the patients in both groups were extracted. The clinical data and imaging features related to MVI were included in univariate analysis and logistic regression analysis to identify the risk factors for MVI in the liver cirrhotic patients with sHCC. A risk prediction model for MVI was constructed, and its predictive value was evaluated using the receiver operating characteristic (ROC) curve, followed by a validation study. Results Among 120 patients included, 34 had MVI, with an incidence rate of 28.33%. Compared with the non-MVI group, the proportions of preoperative serum AFP level >400 μg/L, tumor size ≥5 cm, rough tumor margin, and low enhancement in portal vein phase and delayed phase were higher in those in the MVI group. The clearance time in the MVI group was earlier (P<0.05). Multivariate logistic regression analysis showed that preoperative serum alpha fetoprotein (AFP) level >400 μg/L (OR=4.916), rough tumor margin (OR=4.977), low enhancement in portal vein phase (OR=8.854), and low enhancement in delayed phase (OR=12.455) were independent risk factors for MVI in patients with sHCC under the background of liver cirrhosis, and prolonged clearance time (OR=0.971) was a protective factor (P<0.05). The prediction model was constructed as the follows: P=1/1+[e(3.145-1.096*X1-1.174*X2-1.052*X3-1.058*X4+1.067*X5)] (X1, X2, X3, X4, and X5 corresponding to preoperative serum AFP, tumor margin, enhancement in portal phase, enhancement in delayed phase, and clearance time, respectively). The area under the ROC curve (AUC) was 0.899, and test of goodness of fit showed P=0.854 (>0.05). Internal validation showed that the sensitivity, specificity and accuracy of the model were 62.50% (5/8), 83.33% (10/12) and 75.00% (15/20), respectively. Conclusion The prediction model for MVI in liver cirrhotic patients with sHCC constructed based on gray-scale ultrasound and CEUS features has good discrimination and calibration,thus can be used as a potential non-invasive method for predicting MVI.

Key words: Small hepatocellular carcinoma, Microvascular invasion, Gray-scale ultrasound, Contrast-enhanced ultrasound, Prediction model