A radiomics model based on magnetic resonance imaging to predict cytokeratin 7/19 expression and liver fluke infection of hepatocellular carcinoma

Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer. HCC with liver fluke infection could harbor unique biological behaviors. This study was aimed at investigating radiomics features of HCC with liver fluke infection and establishing a model to predict the expression of cytokeratin 7 (CK7) and cytokeratin 19 (CK19) as well as prognosis at the same time. A total of 134 HCC patients were included. Gadoxetic acid-enhanced magnetic resonance imaging (MRI) images of all patients were acquired. Radiomics features of the tumor were extracted and then data dimensionality was reduced. The radiomics model was established to predict liver fluke infection and the radiomics score (Radscore) was calculated. There were 11 features in the four-phase combined model. The efficiency of the combined model increased significantly compared to each single-phase MRI model. Radscore was an independent predictor of liver fluke infection. It was also significantly different between different expression of CK7/ CK19. Meanwhile, liver fluke infection was associated with CK7/CK19 expression. A cut-off value was set up and all patients were divided into high risk and low risk groups of CK7/CK19 positive expression. Radscore was also an independent predictor of these two biomarkers. Overall survival (OS) and recurrence free survival (RFS) of negative liver fluke infection group were significantly better than the positive group. OS and RFS of negative CK7 and CK19 expression were also better, though not significantly. Positive liver fluke infection and CK19 expression prediction groups harbored significantly worse OS and RFS, survival of positive CK7 expression prediction was unsatisfying as well. A radiomics model was established to predict liver fluke infection among HCC patients. This model could also predict CK7 and CK19 expression. OS and RFS could be foreseen by this model at the same time.