Abstract
Tumor micronecrosis is a validated malignancy biomarker with poor prognosis in hepatocellular carcinoma (HCC), but currently lacks reliable preoperative detection methods. We developed a radiomics model for predicting tumor micronecrosis using preoperative contrast-enhanced computed tomography (CECT). The study included 1017 patients with histopathologically confirmed HCC across three centers, allocated to development, internal test, external test, and application sets. Radiomics features were extracted from tumor regions, with model training via a stratified 5-fold cross-validation framework. The model yield area under the curve values of 0.80 and 0.75 in the internal and external test sets, respectively. Survival curves for histologic and radiomics-predicted micronecrosis groups demonstrated close alignment. For patients treated with tyrosine kinase inhibitor and anti-PD-(L)1 immunotherapy, radiomics-predicted micronecrosis(-) and micronecrosis(+) groups showed comparable survival (P = 0.312). Mass cytometry demonstrated significant correlations between micronecrosis-associated radiomics features and peripheral CD24(+) B cells and neutrophils. The radiomics model shows promise for preoperative micronecrosis assessment.