Abstract
BACKGROUND We aimed to establish a preoperative prediction model of microvascular invasion in hepatocellular carcinoma based on gadoxetic acid disodium-enhanced magnetic resonance imaging in conjunction with serology. MATERIAL AND METHODS In this study, the clinicopathological and imaging data of 120 patients with hepatocellular carcinoma from the Affiliated Huai'an Hospital of Xuzhou Medical University were retrospectively collected, and patients were divided into a training group and a validation group in a ratio of 7: 3. The independent risk factors were determined through logistic regression analysis, and a prediction model was established based on the derived independent risk factors, with imaging as the main factor. Receiver operating characteristic (ROC) curve analysis was performed, and finally, the column line graph, calibration curve, and decision curve of the prediction model were constructed for research. RESULTS Of the 84 patients in the training group, multivariate analysis showed that C-reactive protein to albumin ratio (CAR; OR 35.312, 95% CI 3.064-43.293, P=0.029), tumor signal in hepatobiliary phase (HBP; OR 3.559, 95% CI 1.206-10.499, P=0.021), and peritumoral hypointensity in HBP (OR 15.296, 95% CI 1.741-134.358, P=0.014) were significantly associated with microvascular invasion. When the ROC curve was analyzed for CAR, tumor signal, and peritumoral hypointensity on HBP, the areas under the curve (AUC) of 3-combination applications was 0.789 (95% CI 0.686-0.892). The sensitivity of 3-combination application was 87.5% and specificity was 63.9%. CONCLUSIONS CAR, tumor signal in HBP, and peritumoral hypointensity in HBP are independent risk factors for hepatocellular carcinoma microvascular invasion, and the prediction model constructed by combining the 3 together has better predictive efficacy than that constructed by a single indicator.