Preoperative prediction of early recurrence and microvascular invasion in hepatocellular carcinoma using conventional ultrasound, elastography, and clinical nutritional indicators: a radiomics-based study.

objectivesTo develop and validate preoperative models for predicting 1-year recurrence and microvascular invasion (MVI) in hepatocellular carcinoma (HCC) using conventional ultrasound, elastography, and clinical nutritional indicators.

methodsThis retrospective study included patients with pathologically confirmed HCC who underwent curative hepatectomy and preoperative conventional ultrasound plus elastography. A total of 619 patients were included in the recurrence analysis, and 581 patients with complete pathological MVI assessment and complete modeling variables were included in the MVI analysis. Clinical, laboratory, nutritional, ultrasound, and elastography data were collected. Radiomics features were extracted from grayscale ultrasound and elastography images after standardized preprocessing and region-of-interest segmentation. Clinical, radiomics, and combined models were developed in the training cohorts and evaluated in the validation cohorts.

resultsFor 1-year recurrence prediction, the combined model showed the best performance, with area under the curve (AUC) values of 0.675 in the cross-validation set and 0.771 in the independent test set, compared with 0.567 and 0.656 for the clinical model and 0.660 and 0.759 for the radiomics model, respectively. For MVI prediction, the combined model also achieved the best discriminative performance, with AUC values of 0.860 in the cross-validation set and 0.819 in the independent test set, compared with 0.836 and 0.793 for the clinical model and 0.667 and 0.661 for the radiomics model, respectively.

conclusionsA combined model integrating ultrasound-elastography radiomics and clinical nutritional indicators showed favorable performance for preoperative prediction of both 1-year recurrence and MVI in HCC.