Objective To explore the diagnostic value of CT radiomics combined with prothrombin induced by vitamin K absence or antagonist-Ⅱ(PIVKA-Ⅱ)for hepatocellular carcinoma in the background of liver cirrhosis.Methods The clinical and CT imaging data of patients with liver cirrhosis were analyzed retrospectively.Hepatocellular carcinoma was observed in the observation group,and no hepatocellular carcinoma was found in the control group.Logistic regression was used to analyze the clinical factors of hepatocellular carcinoma and a Clinic model was constructed.Support vector machine(SVM)was used to construct the optimal feature model(Rad model).An artificial neural network model(Combine model)was built based on Softmax policy using Python3.6.Results The degree of liver cirrhosis,aspartate aminotransferase(AST),alpha-fetoprotein(AFP)and PIVKA-Ⅱ were independent factors for predicting hepatocellular carcinoma(P＜0.05).The sensitivity and specificity of the clinical logistic regression model were 73.15％and 68.34％,respectively.The SVM model was used to construct Radiomics score(Radscore)containing 7 optimal features,and there were significant differences between the two groups(P＜0.001).DeLong test showed that the area under the curve(AUC)of Combine model was significantly higher than that of Rad model and Clinic model(P＜0.05).The Hosmer-Lemeshow test showed that the Combine model agrees well.Decision curve analysis showed that the curves of Combine model were significantly higher than Clinic model,Rad model and the extreme curve.Conclusion The Combine model based on CT radiomics combined with clinical factors can accurately predict the occurrence of hepatocellular carcinoma in the background of liver cirrhosis.

Objective To explore the diagnostic value of CT radiomics combined with prothrombin induced by vitamin K absence or antagonist-Ⅱ(PIVKA-Ⅱ)for hepatocellular carcinoma in the background of liver cirrhosis.Methods The clinical and CT imaging data of patients with liver cirrhosis were analyzed retrospectively.Hepatocellular carcinoma was observed in the observation group,and no hepatocellular carcinoma was found in the control group.Logistic regression was used to analyze the clinical factors of hepatocellular carcinoma and a Clinic model was constructed.Support vector machine(SVM)was used to construct the optimal feature model(Rad model).An artificial neural network model(Combine model)was built based on Softmax policy using Python3.6.Results The degree of liver cirrhosis,aspartate aminotransferase(AST),alpha-fetoprotein(AFP)and PIVKA-Ⅱ were independent factors for predicting hepatocellular carcinoma(P＜0.05).The sensitivity and specificity of the clinical logistic regression model were 73.15％and 68.34％,respectively.The SVM model was used to construct Radiomics score(Radscore)containing 7 optimal features,and there were significant differences between the two groups(P＜0.001).DeLong test showed that the area under the curve(AUC)of Combine model was significantly higher than that of Rad model and Clinic model(P＜0.05).The Hosmer-Lemeshow test showed that the Combine model agrees well.Decision curve analysis showed that the curves of Combine model were significantly higher than Clinic model,Rad model and the extreme curve.Conclusion The Combine model based on CT radiomics combined with clinical factors can accurately predict the occurrence of hepatocellular carcinoma in the background of liver cirrhosis.