Prediction Of Early Recurrence Of Hepatocellular Carcinoma After Liver Transplantation Based On CT Radiomics Nomogram

ObjectiveTo explore the value of CT radiomics nomogram in predicting early recurrence in patients with primary hepatocellular carcinoma(HCC)after liver transplantation.MethodsA set of 151 patients with HCC who underwent liver transplantation in our hospital from December 2013 to September 2019 were enrolled in this retrospective study.Patients were randomly divided into a training dataset(105cases)and a validation dataset(46 cases).Recurrence were confirmed by postoperative follow-up.1218 radiomics features were extracted from enhenced CT images of arterial phase.LASSO regression model was used for data dimension reduction.Logistic regression was used to build the prediction model.The clinical factors of patients were collected,and the clinical model was established by univariate analysis and multivariate analysis.A nomogram model was constructed by introducing the radiomics signature into the clinical model.The predictive performance of the radiomics signature,clinical model,nomogram model was evaluated by the area under the curve(AUC)of receiver operating characteristic(ROC).The calibration capability of the nomogram model was evaluated by the calibration curve.Clinical usefulness of the radiomics signature,clinical model and nomogram model was evaluated by the decision curves.ResultsThe selected features extracted from arterial phase with potential predictive value were 13 by applying LASSO regression model,and the radiomics signatures of each patient were calculated based on these 13 radiomics features.After univariate and multivariate analysis,AFP(or = 5.868,95% CI: 1.525-22.585;P = 0.010)and GGT(or = 5.402,95% CI:1.597-18.268;P = 0.003)were included in the clinical model as independent risk factors.By introducing the radiomics signature into the clinical model,the nomogram model was constructed by AFP,GGT and radiomics signature.In the training dataset,the AUC of radiomics signature,clinical model and nomogram model were 0.864,0.747 and 0.882,respectively.while in the validation dataset,the AUC of radiomics signature,clinical model and nomogram model were 0.848,0.735 and 0.917,respectively.In the training dataset,after the radiomics signature was introduced into the clinical model,the predictive performance of the clinical model was significantly improved(from 0.747 to 0.882;P = 0.045).In the validation dataset,after the radiomics signature was introduced into the clinical model,the predictive performance of the clinical model was also significantly improved(from 0.735 to 0.917;P =0.047).In the training dataset and the validation dataset,the accuracy,sensitivity and specificity of the radiomics signature were 84.7%,85.7%,84.5% and 82.6%,91.7%,79.4%,respectively;the accuracy,sensitivity and specificity of the clinical model were 74.2%,76.2%,77.4% and 67.4%,75.0%,64.7%,respectively;the accuracy,sensitivity and specificity of the nomogram model were 91.4%,81.0%,94.0% and 89.1%,91.7%,88.2%,respectively.The calibration curve shows that the prediction probability and the actual probability of the nomogram model in the training dataset and the validation dataset are consistent,indicating that the nomogram model has good calibration capability.Decision curve analysis shows that the nomogram model has better clinical usefulness than the clinical model.Conclusions1.The radiomics signature based on preoperative enhanced CT images of arterial phase is an independent risk factor for early recurrence of HCC patient after liver transplantation.2.The nomogram model based on the integration of clinical model and radiomics signature can significantly improve the predictive performance of clinical model.3.Nomogram model has high calibration capability,which can accurately predict the probability of early recurrence,and provide objective basis for clinicians to make decisions.