摘要
目的探讨人工神经网络模型(Artificial neural network,ANN)在乙型肝炎慢加急性肝衰竭患者发生肝性脑病的危险因素评估中的应用价值。方法 479例乙型肝炎慢加急性肝衰竭患者作为研究对象,采用单因素分析患者的性别、年龄、糖尿病、腹腔感染、肺部感染、上消化道出血、白蛋白、球蛋白、总胆红素、ALT、AST、γ-谷氨酰转肽酶、胆碱酯酶、总胆固醇、血钾、血钠、血肌酐、凝血酶原时间国际标准化比值(INR)、WBC、Hb、血小板、甲胎蛋白、HBV DNA载量并筛选出有统计学意义的临床指标,应用ANN模型和Logistic回归模型进行多因素分析;采用受试者工作特征(ROC)曲线下面积(AUC)评估ANN模型和Logistic回归模型的预测价值。结果 ANN模型提示乙型肝炎慢加急性肝衰竭患者发生肝性脑病的危险因素顺序依次为:INR、WBC、Hb、年龄、血钠、总胆红素、血肌酐、甲胎蛋白、腹腔感染、上消化道出血和肺部感染。Logistic回归模型提示:INR(回归系数0.45,P<0.01,OR值1.566)、WBC(回归系数0.12,P<0.01,OR值1.125)、Hb(回归系数-0.01,P=0.015,OR值0.987)和年龄(回归系数0.02,P=0.026,OR值1.021)为主要危险因素。构建的ANN模型和Logistic回归模型的AUC分别为0.881±0.016、0.825±0.020(Z=2.186,P=0.029)。结论 ANN模型具有较好的危险因素评估预测价值,INR升高、WBC升高、Hb降低和年龄偏大是乙型肝炎慢加急性肝衰竭患者发生肝性脑病的主要危险因素。
Objective To investigate the value of artificial neural network（ANN）in evaluating the risk factors of hepatic encephalopathy in patients with hepatitis B virus-related acute-on-chronic liver failure（HBV-ACLF）.Methods In this retrospective study,479 patients with HBV-ACLF were enrolled.Sex,age,diabetes,celiac infection,pulmonary infection,upper gastrointestinal hemorrhage,albumin,globulin,total bilirubin,alanine aminotransferase（ALT）,aspartate aminotransferase（AST）, gamma-glutamyl transpeptidase, cholinesterase,total cholesterol, potassium, sodium,creatinine,prothrombin time international normalized ratio（INR）,white blood cell（WBC）,hemoglobin（HGB）,platelet（PLT）,alpha-fetoprotein and HBV DNA were analyzed with univariate analysis for screening the significant risk factors for hepatic encephalopathy.Then the ANN model and the logistic regression model were used to determine the risk factors by multivariate analysis.The area under the receiver operating characteristic（ROC）curve（AUC）were used to evaluate the value of ANN model and logistic regression model.Results The ANN model indicated that risk factors of hepatic encephalopathy in patients with HBV-ACLF were INR,WBC,HGB,age,sodium,total bilirubin,creatinine,AFP,celiac infection,upper gastrointestinal bleeding and pulmonary infection sequentially in turn.The logistic regression model showed that INR（B=0.45,P0.001,OR=1.566）,WBC（B=0.12,P0.001,OR=1.125）,HGB（B=-0.01,P=0.015,OR=0.987）,and age（B=0.02,P=0.026,OR=1.021）were the significant risk factors.The AUC of the ANN model and logistic model were 0.881±0.016 and 0.825±0.020（P=0.029）,respectively.Conclusion ANN model might have a powerful clinic value in evaluating risk factors of hepatic encephalopathy in patients with HBV-ACLF,which revealed that higher INR,higher WBC,lower HGB,and older age were the significant risk factors of hepatic encephalopathy.
出处
《肝脏》
2017年第12期1085-1089,1093,共6页
Chinese Hepatology
基金
福州市卫生系统科技计划项目(2014-S-W26)
关键词
肝功能衰竭
肝性脑病
危险因素
人工神经网络
回归分析
Liver failure
Hepatic encephalopathy
Risk factors
A rtificial neural network （A NN）
Regression analysis
