Clinical Value Of Radiomics Based On CT In Predicting Tumor Recurrence After Liver Transplantation For Hepatocellular Carcinoma

Objectives: The high-throughput features of valuable HCC imaging ensembles were explored by radiomics to explore whether these features were associated with recurrence of HCC after liver transplantation.An attempt was made to establish an radiomics prediction model for predicting recurrence of HCC liver transplantation.The aim is to predict the risk of postoperative recurrence by imaging before surgery,and then to risk stratification of HCC patients,to make reasonable clinical decisions,and to guide clinical organ allocation.Methods: 1.A total of 133 patients who underwent classic orthotopic liver transplantation in the First Central Hospital of Tianjin from October 2011 to December 2016 were selected and pathologically proved to be HCC.2.The enrolled patients were randomly divided into training set and verification set according to 7:3,93 training sets and 40 verification sets.The training set was used to establish the predictive model,and the verification set was used to verify the prediction performance of predictive model.3.HCC lesions in patients were delineated with in four-phase of enhanced CT images,and screen and analyze stable radiomics features using minimum absolute contraction and selection algorithm(LASSO)and Cox regression model.4.To explore the correlation between selected stable radiomics features and HCC recurrence after liver transplantation.An radiomics prediction model was established by using the stable radiomics features in four phases of enhanced CT images respectively,and the prediction performance of each phase prediction model was compared.5.To select and evaluate the clinical laboratory indicators related to HCC recurrence after liver Transplantation,and to establish a clinical model to predict HCC recurrence after liver transplantation.6.Finally,combining with the radiomics features of HCC patients and effective clinical laboratory indicators,a joint prediction model for predicting HCC recurrence after liver transplantation was established.7.To compare the predictive performance of three HCC recurrence prediction models after liver transplantation,and to establish an individualized clinical prediction program for HCC recurrence after liver transplantation.Results: 1.By delineating the radiomics features of HCC lesions in the fourth stage of enhanced CT,the high-throughput features of the imaging related to tumor recurrence after liver transplantation were extracted and screened.No stable radiomics features correlatded with HCC recurrence was found in the delayed and balanced periods.The stable radiomics features were selected to establish a corresponding prediction model related to HCC recurrence.By comparison,the prediction model of tumor recurrence after liver transplantation based on the HCC stable radiomics features extracted from arterial phase has a better predictive value in the the fourth phase of enhanced CT images.The number of original features found in the enhanced CT arterial phase was 853,and the stable radiomics features was 84.Finally,9 stable radiomics features with significant correlation with HCC recurrence after liver transplantation were screened from 84 candidate features of arterial phase(p < 0.001).The established radiomics predictive model for predicting recurrence of HCC after liver transplantation has a predictive index C-index of 0.743(95%CI: 0.632~0.853)in training set and 0.705(95%CI:0.537~0.874)in the verification set.2.Through single factor analysis and multivariate analysis,HBs Ag(HR=0.255,95%CI: 0.099~0.654;P=0.004)and BCLC(HR=1.373,95%CI: 1.086~1.735)were found effective predictors of HCC recurrence after liver transplantation.The established clinical predictive model for predicting recurrence of HCC after liver transplantation has a predictive index C-index of 0.675(95%CI: 0.568-0.782)in the training set and 0.713(95%CI:0.549~0.877)in the validation set.3.The comprehensive predictive model consisting of imaging omics and clinical risk factors has the best predictive performance among the three predictive models.The predicted index C-index is 0.785(95% confidence interval [ CI]: 0.674~0.895)in the training set,0.789(95%CI: 0.620~0.957)in validation set.The calibration curve was shown to be substantially consistent with the actual recurrence in the training set(P=0.121)and the validation set(P=0.164).4.Three predictive models for predicting HCC recurrence after liver transplantation were established.They were used to predict the early recurrence of HCC after liver transplantation in patients within or exceeding Milan standard.The comprehensive predictive model has the best predictive performance,and its C-index is 0.773(95%CI: 0.532-1.000)in patients within Milan standard,and 0.726(95%CI: 0.623-0.829)in patients exceeding Milan standard.Conclusions: 1.High-throughput features of radiomics in CT imaging can be used as important predictors of liver transplantation recurrence of HCC.2.Combination of radiomics and traditional clinical parameters can improve the predictive efficiency of HCC recurrence after liver transplantation.3.Radiomics can predict the risk of tumor recurrence in patients with HCC after liver transplantation before surgery,and then make reasonable clinical decisions and guide clinical practice.4.The comprehensive predictive model can be used as a reference and supplement for the Milan standard of liver transplantation for HCC patients.