Establishment Of Prognostic Prediction Model Of Stent Implantation In Malignant Biliary Obstruction And Evaluation Of Artificial Neural Network

Part ? Prediction model of early biliary tract infection after stent implantation for malignant biliary obstruction Establishment and Evaluation of Artificial Neural NetworkObjective To construct a predictive model for early biliary infection(EBI)after endoscopic retrograde cholangiopancreatography(ERCP)combined with biliary stent implantation in patients with malignant biliary obstruction(MBO),and to evaluate the predictive performance of the model.Methods The clinical data of patients who underwent ERCP combined with biliary stent implantation in the Department of Hepatobiliary Surgery,Hospital of Ningxia Medical University from January 2018 to September 2021 were retrospectively collected to observe whether early biliary tract infection occurred after operation(within 30 days).All enrolled patients were divided into training group and validation group.The data of training group were analyzed by univariate and multivariate analysis to obtain independent risk factors affecting postoperative EBI and construct a logistic prediction model.The model was verified by internal and external data,and the receiver operating characteristic(ROC)curve and calibration curve were drawn.The area under the curve(AUC)and H-L test values were calculated to evaluate the predictive ability of the model.Importance of using artificial neural network(ANN)to assess independent risk factors.Results A total of 285 patients were included in the study,200 in the training group and85 in the validation group.Univariate and multivariate analysis showed that the independent risk factor for EBI was obstruction location(OR,5.942;95 % CI,2.507-14.081;p < 0.001),with or without gallstones(OR,4.821;95 % CI,2.087-11.138;p < 0.001),diabetes mellitus(OR,5.407;95 % CI,2.067-14.148;p = 0.001),obstruction length(OR,1.058;95 % CI,1.028-1.089;p < 0.001),so as to construct a logistic regression model and visualize it with a nomogram.The AUC values were 0.807 and 0.831,and the H-L test P values were 0.845 and0.197,respectively.The model had strong predictive ability and good goodness of fit.The best critical risk value of the model was 0.225.ANN evaluation suggested that obstruction length was the most important predictor variable.Conclusion The logistic model based on the preoperative clinical data of MBO patients has a good predictive ability for the risk of postoperative EBI,and ANN shows that the length of obstruction is the most important predictor.Part II Establishment and application of early warning model for early death after stent implantation in malignant biliary obstruction: Comparison of logistic model and ANN modelObjective To construct an early warning model for early death(within 30 days)after ERCP combined with biliary stent implantation in MBO patients and evaluate the predictive ability of logistic regression model and ANN model.Methods The clinical data of patients undergoing ERCP combined with biliary stent implantation in the Department of Hepatobiliary Surgery,Hospital of Ningxia Medical University from January 2018 to September 2021 were retrospectively collected to observe whether there was early death(within 30 days)after operation.All enrolled patients were divided into training group and validation group.The potential risk factors of death within 30 days were obtained by univariate analysis in the training group,and the ANN model was constructed.The ROC curve was drawn and the sensitivity,specificity and accuracy of the model prediction were calculated.Logistic model was constructed by multivariate analysis of independent risk factors,ROC curve and calibration curve were drawn,and AUC value and H-L goodness of fit test value were calculated.The sensitivity,specificity and accuracy of the model were calculated by the training group and the verification group.Results Univariate analysis showed that the potential risk factors were CA199,ERCP history,liver metastasis,CEA and NLR.Multivariate analysis showed that the independent risk factors were CA199 and ERCP history.logistic model was constructed for independent risk factors and ANN model was constructed for potential risk factors.The ROC curves of the logistic model and the ANN model showed that the ANN model had a larger AUC value(ANN model: 0.814,logistic model: 0.727).The accuracy(ANN model: 84.7 ～ 86.0 %,logistic model: 59.0 ～ 65.0 %)and specificity(ANN model: 92.5 ～ 93.8 %,logistic model:76.9 ～ 83.3 %)of the ANN model were also higher,and there was no significant difference in sensitivity between the two models,indicating that the ANN model had better predictive performance.Conclusion ANN model is superior to logistic model in predicting the risk of early death after ERCP combined with biliary stent implantation in MBO patients.