Objective To investigate the value of dynamic contrast-enhanced magnetic resonance imaging (MRI) derived intratumoral and peritumoral imaging biomarker models in predicting the positive expression of cytokeratin 19 (CK19) in hepatocellular carcinoma (HCC).

Methods Dynamic enhanced MRI images from 120 patients with HCC were retrospectively collected. Using stratified random sampling, the cases were divided into training set (84 cases) and validation set (36 cases) at a ratio of 7∶3. Radiomic features were extracted from the intratumoral region, as well as the 2 mm and 4 mm peritumoral regions, during the arterial, portal venous and delayed phases. Feature dimensionality reduction and selection were performed using the t-test or Mann-Whitney U test, Pearson correlation analysis, maximum relevance-minimum redundancy (mRMR) and LASSO regression. Logistic regression (LR) algorithms were employed to construct radiomics models for the intratumoral region, 2 mm peritumoral area, 4 mm peritumoral area, and combined model of intratumoral and optimal 2 mm peritumoral regions. The performance of each model was evaluated using receiver operating characteristic (ROC) curves and decision curve analysis (DCA).

Results In the validation set, the areas under the curve (AUC) for the intratumoral, 2 mm peritumoral, and 4 mm peritumoral regions were 0.840, 0.760 and 0.742, respectively. The combined model of intratumoral and optimal tumor (2 mm peritumoral) regions achieved an AUC of 0.976 in the training set and 0.869 in the validation set.

Conclusion The 2 mm model in the intratumoral and peritumoral regions based on dynamic contrast-enhanced MRI can effectively predict the expression of CK19 in hepatocellular carcinoma, providing a feasible method for non-invasive clinical detection of CK19 status.