Related Experiment Video
Updated: May 20, 2026
Non-Invasive PET/MR Imaging in an Orthotopic Mouse Model of Hepatocellular Carcinoma
Published on: August 31, 2022
Radiomics models to predict proliferative small hepatocellular carcinoma and its prognosis based on Gd-EOB-DTPA
Fengxi Chen1, Xuefeng Li2, Dandan Wang3
17T Magnetic Resonance Translational Medicine Research Center, Department of Radiology, Southwest Hospital, Army Medical University (Third Military Medical University), Chongqing 400038, China; Department of Radiology, The Second Affiliated Hospital of Chongqing Medical University, Chongqing 400010, China.
Objective:
To develop and validate radiomics models based on gadoxetate disodium (Gd-EOB-DTPA)-enhanced MRI for preoperative prediction of proliferative small hepatocellular carcinoma (sHCC) and to evaluate their prognostic value.
Methods:
A total of 331 patients with pathologically confirmed sHCC from two institutions were retrospectively enrolled and divided into training (n = 147), internal validation (n = 63), external validation (n = 59), and exploratory validation (n = 62) cohorts. Independent predictors were identified using multivariate logistic regression to construct a clinical-radiological model. Intratumoral and peritumoral volumes of interest (VOIs) were manually delineated on hepatobiliary phase (HBP) images from Gd-EOB-DTPA-enhanced MRI, followed by feature extraction and selection to develop intratumoral, peritumoral, and combined radiomics models. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), interpretability was analyzed using Shapley additive explanations (SHAP), and recurrence-free survival (RFS) was assessed with the Kaplan-Meier method.
Results:
In the external validation cohort, the combined radiomics model achieved the best predictive performance (AUC = 0.836), outperforming both the intratumoral (AUC = 0.748, P = 0.049) and peritumoral (AUC = 0.744, P = 0.044) models, with consistent results in the exploratory validation cohort (AUC = 0.817). No significant difference was observed between the intratumoral and peritumoral models (P = 0.863). All radiomics models outperformed the clinical-radiological model (P < 0.001). SHAP analysis identified original_shape_Elongation_Intra, log-sigma-2-0-mm-3D_firstorder_Maximum_Intra, and wavelet-HLH_firstorder_Kurtosis_Intra as the most influential features. Patients with proliferative sHCC showed significantly shorter RFS than those with non-proliferative sHCC (P = 0.049), and high-risk patients identified by the combined radiomics model exhibited poorer RFS (P = 0.025).
Conclusion:
The combined radiomics model provides a promising and noninvasive approach for preoperative prediction of proliferative sHCC and enables effective postoperative risk stratification, offering potential value for individualized clinical management.
