Abstract: Objective To investigate the correlation between MRI enhancement features, radiomics characteristics and PFKFB3 expression in primary hepatocellular carcinoma(HCC)tissue, and to establish a radiomics prediction model for drug-resistant related proteins of HCC. Methods Information of 135 HCC patients who received preoperative multiphase MRI and surgical resection during Jan. 2015 and Dec. 2020 was retrospectively analyzed. The clinical data(age, gender, history of smoking and drinking, alanine aminotransferase, aspartate transaminase, alpha-fetoprotein, pathologic stage and hepatitis B infection), conventional imaging features(tumor size, capsular, enhancement characteristics in arterial phase, necrosis, portal vein invasion, blood-supply type, hemorrhage, intrahepatic satellite foci and arterial tumor-liver differences in arterial phase), and radiomic features were recorded. The expression of PFKFB3 was detected with immunohistochemistry. The independent predictors were screened with multivariate analysis(P<0.05). The radiomics prediction model was constructed based on the features of the selected training set. The receiver operating characteristic(ROC)curve was drawn. The accuracy of the prediction model was evaluated with the area under the curve(AUC)and verified in the validation set. Results Alanine aminotransferase(OR=0.36, 95%CI:0.16-0.83, P=0.017)and the presence of intrahepatic satellite foci(OR=6.89, 95%CI:1.76-27.03, P=0.006)were independent predictors of positive PFKFB3 expression. The AUC of the MRI radiomics model was 0.99 in the training set and 0.80 in the validation set, with 95%CI of 0.61-1.00, sensitivity of 0.78 and specificity of 0.75. Conclusion The model of enhanced MRI radiomics can predict the expression of PFKFB3 in primary HCC, which can provide important information of tumor drug resistance in the treatment of HCC.

Key words: Hepatocellular carcinoma, Drug resistance, PFKFB3, Radiomics, Magnetic resonance imaging