基于高频超声定量参数及临床相关资料构建早产儿主动脉瓣二瓣化畸形发病预测模型

Construction of the Early Warning Model of Aortic Valve Dicuvalization in Premature Infants Based on the Quantitative Parameters of High Frequency Ultrasound and Clinical Data

目的 基于高频超声定量参数及临床相关资料构建早产儿主动脉瓣二瓣化畸形(BAV)发病预测模型。方法 选取2020年1月—2022年12月收治的86例BAV早产儿作为研究组,按照1︰1比例选取86例无BAV的早产儿作为对照组。统计2组临床资料、高频超声定量参数及表现,构建随机森林模型和多因素Logistic回归方程模型,采用受试者工作特征(ROC)曲线分析2种模型预测早产儿BAV发病的效能。结果 2组早产儿25-羟基维生素D、心肌型脂肪酸结合蛋白(H-FABP)、心肌肌钙蛋白I(cTnI)、心肌型肌酸激酶同工酶、家族史、高频超声主动脉瓣正向最大血流速度(V_(max))、高频超声跨瓣压差、高频超声主动脉瓣狭窄、高频超声主动脉瓣反流、高频超声主动脉受累扩张及早产儿母亲年龄比较差异有统计学意义(P<0.01)。采用随机森林算法构建早产儿BAV发病预警模型,根据平均准确度下降程度进行重要性排序,从高到低依次为早产儿高频超声V_(max)、高频超声跨瓣压差、高频超声主动脉瓣狭窄、H-FABP、cTnI、家族史,该模型预测早产儿BAV发病的曲线下面积(AUC)为0.974,敏感度为0.954,特异度为0.988;多因素Logistic回归方程显示,早产儿高频超声V_(max)、高频超声跨瓣压差、高频超声主动脉狭窄及H-FABP、家族史、cTnI是早产儿BAV发病的影响因素(P<0.01),据此构建的多因素Logistic回归方程预测模型预测早产儿BAV发病的AUC为0.924,敏感度为0.919,特异度为0.930。结论 基于高频超声定量参数及临床相关资料构建早产儿BAV发病预警模型具有可行性,且预测效能良好,能为临床诊疗工作提供参考。

Objective To establish an early warning model of bicuspid aortic valve(BAV)malformation in premature infants based on quantitative parameters of high frequency ultrasound and clinical data.Methods A total of 86 premature infants with BAV treated from January 2020 to December 2022 were selected as the research group,and 86 preterm infants without BAV were selected as the control group according to a ratio of 1︰1.The clinical data,quantitative parameters of high frequency ultrasound and manifestations of the two groups were analyzed.The random forest model and multivariate Logistic regression equation model were constructed,and the efficacy of the two models in predicting the onset of BAV in premature infants was analyzed by receiver operating characteristic(ROC)curve.Results 25 hydroxyvitamin D,heart-type fatty acid binding protein(H-FABP),cardiac troponin I(cTnI),myocardial creatine kinase isoenzyme,family history,maximum blood flow velocity(V max)of the aortic valve by high frequency ultrasound,cross-valve pressure difference by high frequency ultrasound,aortic stenosis by high frequency ultrasound,aortic regregence by high frequency ultrasound,aortic involvement dilatation by high frequency ultrasound and age of mothers of preterm infants in the two groups were statistically significant(P<0.01).The early-warning model of BAV in premature infants was constructed using the random forest algorithm,and the importance was ranked according to the average accuracy decline,including Vmax by high frequency ultrasound,cross-valve pressure difference by high frequency ultrasound,aortic stenosis by high frequency ultrasound,H-FABP,cTnI,and family history from high to low in the preterm infants.The area under the curve(AUC),sensitivity and specificity of the model in predicting BAV in preterm infants were 0.974,0.954 and 0.988,respectively.Multivariate Logistic regression equation showed that V max by high frequency ultrasound,cross-valve pressure difference by high frequency ultrasound,aortic stenosis by high frequency ultrasound,H-FABP,family history and cTnI were the influencing factors for the incidence of BAV in preterm infants(P<0.01).The multivariate Logistic regression model predicted that the AUC,sensitivity and specificity of the model in predicting BAV in preterm infants were 0.924,0.919 and 0.930,respectively.Conclusion It is feasible to construct an early warning model of premature BAV based on quantitative parameters by high-frequency ultrasound and clinical data,and the prediction efficiency is good,which can provide convenience for clinical diagnosis and treatment.