Predicting microvascular invasion in hepatocellular carcinoma using US features combined with clinicopathological characteristics: a multicenter study

PURPOSE: To enhance the clinical management of hepatocellular carcinoma (HCC), we explored how conventional ultrasound and pathological features vary with microvascular invasion (MVI) status. We identified key risk factors for MVI and developed a straightforward preoperative model to predict MVI status. METHODS: In this study, we enrolled 362 patients with HCC confirmed by surgical pathology from three medical institutions, collecting their clinical pathological and ultrasound features. Patients from institutions 1 and 2 were randomly split into training and internal validation sets at a 7:3 ratio, while those from institution 3 served as the external validation set. We used least absolute shrinkage and selection operator (LASSO) and best subset regression (BSR) to select variables for model building. A nomogram was constructed to predict hepatocellular carcinoma MVI, and its accuracy, clinical utility, and discrimination were assessed using ROC, clinical decision curves, and calibration plots. RESULTS: A total of 175 patients were included in the study. Of these, 135 patients from institutions 1 and 2 were randomly divided into a training set (95 patients) and an internal validation set (40 patients). Among them, 57.0% (77/135) were MVI-negative, and 43.0% (58/135) were MVI-positive. Only total cholesterol (TC) levels were significantly higher in the MVI-positive group (p