Key points are not available for this paper at this time.
Abstract Purpose: To investigate the value of radiomics in differentiating combined hepatocellular-cholangiocarcinoma (cHCC-CCA) and hepatocellular carcinoma (HCC). Materials and Methods: We retrospectively collected the clinical, serum biomarkers and imaging data of cHCC-CCA (n = 42) and HCC (n = 117) patients. The optimal radiomics features were extracted from CT plain scan, arterial phase, venous phase and delayed phase images for constructing radiomics models. The clinical model, radiomics model and fusion model were constructed by extreme gradient boosting (XGB), and the models were validated with an independent validation cohort. Area under curve (AUC), specificity, sensitivity and decision curve analysis (DCA) were used to evaluate the model efficacy. Results: The fusion model based on CT radiomics performed the best, with an AUC of 0.969, which was superior to the clinical model (AUC=0.860) and the CT radiomics model (AUC=0.853). DCA showed that the fusion model had the highest clinical net yield compared to the other two models. Conclusion: The fusion model based on CT radiomics has good performance in distinguishing cHCC-CCA from HCC, which can better assist in individualized clinical decision-making for patients with cHCC-CCA.
Xu et al. (Mon,) studied this question.