Bayes判别分析在预测剖宫产术后子宫瘢痕妊娠患者结扎子宫动脉的应用

Preliminary study in predicting the uterine artery occlusion by using Bayesian discriminant analysis in cesarean scar pregnancy

目的探讨Bayes判别分析用于瘢痕妊娠患者(CSP)结扎子宫动脉风险预测的临床效果。方法收集2018年6月至2020年10月深圳市中西医结合医院73例未结扎与87例结扎子宫动脉的CSP患者,纳入手术有关因素,通过单因素分析筛选差异变量,采用logistic回归分析模型进行回归分析;利用受试者工作特征(ROC)曲线分析指标预测结扎子宫动脉的价值,将选取变量带入Bayes判别模型中得到判别函数并对其进行检验。结果最终纳入判别方程的变量为:孕囊与瘢痕关系(a1),子宫瘢痕处肌层厚度(a2),瘢痕血流(a3),建立的Bayes判别函数:Y1=-7.138+3.259×a1+2.879×a2+0.035×a3;Y2=-7.239+5.169×a1+1.732×a2+2.166×a3;当Y1>Y2时归为不结扎子宫动脉;当Y1>Y2时归为不结扎子宫动脉;当Y1<Y2时归为结扎子宫动脉,对判别函数进行检验,结果内外部符合率均达到80%以上。结论Bayes判别方程可从统计学角度对CSP患者结扎子宫动脉进行进行初步预测。

Objective To investigate the clinical effect of Bayesian discriminant analysis in predicting the risk of uterine artery occlusion in cesarean scar pregnancy(CSP)patients.Methods According to the unified inclusion and exclusion criteria of this study,73 cases without uterine artery occlusion and 87 cases with uterine artery occlusion of CSP patients were enrolled from Shenzhen Hospital of Integrated Traditional Chinese and Western Medicine.Differential variables were identified based on the results of univariate analysis.Logistic regression analysis model was used for regression analysis.The ROC curve analysis was used to analyze the discriminant indexes in predicting uterine artery occlusion.The discriminant indexes were put into the Bayesian discriminant model to obtain the Bayesian discriminant function.The discriminant function was the retrospectively examined and externally tested.Results The variables finally substituted in the Bayes formulas were the position between gestational sac and scar(a1),the thickness of scar(a2),and the blood flow of cesarean scar(a3)detected by ultrasound.The established Bayes discriminant function was Y1=-7.138+3.259×a1+2.879×a2+0.035×a3;and Y2=-7.239+5.169×a1+1.732×a2+2.166×a3.After putting the aforementioned variables(a1-a3)of CSP patient into the formulas,if the value of Y1 was greater than that of Y2,then the patient was determined as without uterine artery occluded,otherwise the patient was determined as uterine artery occluded.The internal and external conformity rates of the Bayesian discriminant model were all more than 80%.Conclusion Our Bayes formulas can be used to predict the uterine artery occluding for scar pregnancy treatment.