The Value Of ¹⁸F-FDG PET/CT Metabolic Parameters And Radiomics In Predicting Early Recurrence Of Hepatocellular Carcinoma After Liver Transplantation

Objective:To construct a clinical nomogram based on ¹⁸F-FDG PET/CT metabolic parameters to predict early recurrence of hepatocellular carcinoma（HCC）after liver transplantation.Further,an radiomics nomogram combining ¹⁸F-FDG PET/CT radiomics and clinical factors was established to evaluate the value in predicting early recurrence of HCC after liver transplantation.Methods:In part one,62 patients with HCC who underwent liver transplantation in the Affiliated Hospital of Qingdao University from January 2010 to June 2020 were enrolled.These patients underwent PET/CT examination within 3 weeks before liver transplantation.And the follow-up time for patients without recurrence after liver transplantation was at least 1 year.The following information were collected by searching the medical records.Basic clinical factors included age,gender,body mass index,performance status（PS）,basic liver diseases,status of cirrhosis,operation interval,pretreatment before liver transplantation,model for end-stage liver disease（MELD）score,Child-Pugh stage,Barcelona Clinic Liver Cancer（BCLC）staging system,Milan criteria（within or beyond）,UCSF criteria（within or beyond）,and the use of immunosuppressive agents after liver transplantation.Pathology factors included the Edmondson-Steiner stage,the maximum diameter of the tumour in the pathological specimen（Pathdmax）,liver capsule invasion,microvascular invasion（MVI）,nerve invasion,satellite nodule,bile duct invasion,portal tumour thrombus,and donor liver steatosis.Laboratory factors included alpha fetoprotein（AFP）,platelet count（PLT）,alanine aminotransferase（ALT）,aspartate aminotransferase（AST）,total bilirubin（TB）,conjugated bilirubin（CB）,albumin（ALB）,total bile acid（TBA）,serum creatinine（Scr）,prothrombin time（PT）,and international normalized ratio（INR）.CT features included tumor shape,boundary,density,location,subtype,hemorrhage or necrosis,the maximum cross-sectional diameter of the tumour in CT（CTdmax）,and the number of tumours.PET metabolic parameters included the maximum standardized uptake value（SUVmax）and the total lesion glycolysis（TLG）of normal body weight（BW）and lean body mass（LBM）,metabolic tumour volume（MTV）and tumour-to-normal liver uptake ratio（TLR）.Univariate Cox hazard analysis and Multivariate Cox regression analysis were used to select one best predictor from each group above to construct nomograms.Multivariate Cox regression analysis was used to establish a nomogram combining basic clinical data,pathology data,laboratory data,and CT features（CPLC）without metabolic parameters and a nomogram combining basic clinical data,pathology data,laboratory data,CT features,and PET metabolic parameters（CPLCP）with metabolic parameters.The better clinical nomogram was determined by calculating the net reclassification index（NRI）and the integrated discrimination improvement（IDI）between the nomograms.The time-dependent area under the receiver operating characteristic curve（time-AUC）and the concordance index（C-index）were used to compare the efficacy of the better clinical nomogram,Milan criteria and UCSF criteria in predicting early recurrence after liver transplantation in HCC patients.The decision curve analysis（DCA）was used to evaluate the clinical practicability of each model.The Bootstrap method was used for internal verification.In part two,The regions of interest（ROI）in CT images and PET images of 62 patients were drawn and the radiomic features were extracted.The optimal radiomic features were selected by inter-and intra-class correlation coefficients（ICCs）and Lasso-Cox regression.Then the radiomic scores（Rad-score）were calculated.Combined Rad-score and clinical predictors,an radiomics nomogram（R-CPLCP nomogram）were established by using multivariate Cox regression analysis.Time-AUC and C-index were used to compare the efficacy of clinical nomogram and radiomics nomogram in predicting early recurrence after liver transplantation in HCC patients.DCA was used to evaluate the clinical practicability of clinical nomogram and radiomics nomogram,and Bootstrap method was used for internal verification.Results:In part one,Thirty out of 62 patients experienced a recurrence during the one-year follow-up.BCLC stage（P=0.009）,MVI（P=0.032）,AFP（P=0.004）,CTdmax（P=0.033）,and MTV（P=0.039）were the clinical predictors.The CPLC nomogram and the CPLCP nomogram were established.Compared with the CPLC nomogram,the NRI of the CPLCP nomogram increased by 38.98%（95%CI=-18.77-60.43%）and the IDI increased by 4.40%（95%CI=-1.00-16.62%）.The C-index of the CPLCP nomogram was0.774,which was higher than the CPLC nomogram（C-index=0.768）.The CPLCP nomogram was the better clinical nomogram.The AUC value of the CPLCP nomogram was higher than those of Milan criteria and UCSF criteria in the time-AUC.The C-index of the CPLCP nomogram was higher than Milan criteria（C-index=0.637）and UCSF criteria（C-index=0.631）.The DCA curve showed that clinical practicability of the CPLCP nomogram outperformed the Milan criteria and UCSF criteria.In part two,a total of 5 optimal radiomic features were obtained,and the radiomics nomogram（R-CPLCP nomogram）were established.The time-AUC curve showed that at most time points,the AUC values of the R-CPLCP nomogram were higher than that of the CPLCP nomogram.The C-index value of R-CPLCP nomogram was 0.800,which was higher than that of CPLCP nomogram（C-index=0.774）.The The DCA curve showed that within most of the threshold probability range,the net benefit of the R-CPLCP nomogram was higher than that of CPLCP nomogram.The Bootstrap method showed that the prediction efficiency of the R-CPLCP nomogram was higher than that of the CPLCP nomogram.Conclusion:The clinical nomogram（CPLCP nomogram）combining basic clinical factor（BCLC stage）,pathology factor（MVI）,laboratory factor（AFP）,CT feature（CTdmax）,and PET metabolic parameter（MTV）showed good efficacy and high clinical practicability in predicting early recurrence after liver transplantation in HCC patients.Radiomics nomogram（R-CPLCP nomogram）can further improve the prediction efficiency and obtain higher clinical net benefits.