MRI-based radiomics predicts complete responses to initial transcatheter arterial chemoembolization in virus-associated early and intermediate stage HCC

Abstract

Background

To evaluate the value of a multiparametric MRI-based nomogram on predicting response to transcatheter arterial chemoembolization (TACE) in virus-associated hepatocellular carcinoma (HCC) patients;

Methods

This study enrolled 235 and 51 patients from Center 1 and 2, respectively. All patients underwent baseline MRI scans before treatment. The least absolute shrinkage and selection operator (LASSO) regression method was used to screen radiomics features from intra- and peri-tumor areas to establish the radiomics signatures (RS). The nomogram model was built by integrating the RS and clinical predictors. Receiver operating characteristics (ROC), calibration and decision curve analyses (DCA) curves were used to assess predictive performances of radiomics models;

Results

In the training, internal validation and external validation cohort, the AUCs based on developed RS were 0.848, 0.759 and 0.762, respectively. The clinical model consisted of 4 significant distinct clinical predictors, including HBsAg, AFP, BCLC staging and size. To enhance diagnostic efficiency, we integrated 11 radiomics features and 4 clinical predictors to develop a nomogram model, which showed increased AUCs to 0.892, 0.851 and 0.787 in the training, internal validation and external validation cohort, respectively;

Conclusions

This study demonstrates that multiparametric MRI-based radiomics nomogram can preoperatively predict responses to TACE in virus-associated HCC.