The Value of Predictive Models Based on CT and MRI Enhanced Portal Phase Imaging and Clinical Indicators in Predicting Early Recurrence After Resection of Single Hepatocellular Carcinoma
Objective To explore the value of predicting early recurrence(≤2 years)after resection of single hepato-cellular carcinoma(HCC)based on CT and MRI enhanced portal phase imaging and clinical index prediction models.Methods From April 2016 to February 2020,88 patients with single episode HCC who underwent general surgical resec-tion and were pathologically confirmed at Xuzhou Central Hospital were retrospectively collected.Among them,40 patients with early postoperative recurrence were randomly divided into a training set(n=70)and a validation set(n=18)based on a 4∶1 ratio.Manually segment the entire tumor on CT and MRI enhanced portal phase images to obtain regions of inter-est and extract imaging features.In the training set,the univariate selection method,least absolute shrinkage and selection operator(LASSO),and Spearman correlation analysis(Spearman)were used to reduce the dimensions of the histological features and select the best feature set.Single factor and multiple factor analysis was used to determine the high-risk clini-cal factors for early postoperative recurrence in HCC patients.The CT clinical model,MR clinical model,and CT MR clini-cal model were constructed by combining imaging characteristics and clinical high-risk factors.The prediction value of the model was evaluated using the area under the subject working characteristic(ROC)curve(AUC)and Delong test,and a visual nomogram of the prediction model was generated.Calibration curves were used to evaluate the efficacy of the nomo-gram,and decision curve analysis(DCA)was used to evaluate the clinical application value of the nomogram.Results Clinical univariate and multivariate analysis showed that aspartate aminotransferase(AST)was an independent risk factor for predicting early recurrence of HCC(P＜0.05).Based on CT and MRI enhanced portal phase,a total of 9 CT imaging features and 4 MRI optimal imaging features were selected,and a CT clinical model,an MR clinical model,and a CT MR clinical model were constructed respectively with clinical independent risk factors.The area under the curve(AUC)of the ROC curve in the training set and the validation set were 0.919,0.959,0.971,and 0.844,0.850,and 0.875,respectively.The CT-clinical model was significantly better than the MR-clinical model in predicting early postoperative recurrence of HCC(P＜0.05),and the prediction efficiency of the CT-MR-clinical model in the training set was significantly better than the CT-clinical model(P＜0.05).DCA showed that the threshold probability of the training set was＞30％,and the clini-cal net income of the CT-MR-clinical model was higher.Conclusion The prediction model based on CT and MRI en-hanced portal phase imaging features combined with clinical factors AST(CT MR clinical model)has better predictive effi-cacy in predicting early postoperative recurrence in HCC patients,and the prediction efficacy of CT clinical model is superi-or to that of MR clinical model.
NETL
NSTL
万方数据
