基于人工智能的围生期孕妇宫颈分泌物B族链球菌阳性影响因素筛选及预测模型构建  

Screening of influencing factors and construction of a prediction model for group B streptococcus infection in perinatal women based on artificial intelligence

作　　者：袁志平[1] 傅鹰[3] 董佳琳 来汉江[5] 朱水龙[6] 戴军[2] YUAN Zhiping;FU Ying;DONG Jialin;LAI Hanjiang;ZHU Shuilong;DAI Jun(Department of Laboratory,the Second People's Hospital of Xiaoshan District,Hangzhou 311241,China;不详)

机构地区：[1]杭州市萧山区第二人民医院检验科,311241 [2]杭州市萧山区第二人民医院妇产科,311241 [3]浙江大学医学院附属邵逸夫医院检验科 [4]杭州市第九人民医院检验科 [5]杭州市萧山区第一人民医院检验科 [6]杭州市萧山区第三人民医院检验科

出　　处：《浙江医学》2025年第4期382-386,402,共6页Zhejiang Medical Journal

基　　金：萧山区科学技术协会2021年度重点软课题(202105)。

摘　　要：目的探讨围生期孕妇宫颈分泌物B族链球菌(GBS)阳性的影响因素,构建GBS阳性的预测模型,为改善妊娠结局提供更多参考。方法回顾性选取2021年1至7月杭州市萧山区第一、二和三人民医院产科门诊及病房的1538例围生期孕妇为研究对象,根据宫颈分泌物GBS培养结果分为GBS阳性组和GBS阴性组,收集两组孕妇的临床资料并进行比较,再采用7种人工智能算法[逻辑回归、高斯朴素贝叶斯(GNB)、多层感知机、支持向量机、随机森林、梯度增强算法和决策树]来构建GBS阳性的预测模型。模型训练集和验证集按照8∶2的比例随机分组。通过计算ROC曲线的AUC、灵敏度、特异度等指标来测试模型的效能,并对菌株进行药敏试验。结果1538例孕妇中GBS阳性98例,阳性率为6.37%。单因素分析显示,孕妇年龄、孕次、孕前检查、流产史、妊娠期阴道炎及妊娠期糖尿病等临床指标均与GBS阳性有关(均P<0.05)。在GNB算法下,6项临床指标组合的GBS阳性预测模型具有最优的诊断效能(AUC=0.800,灵敏度为0.675,特异度为0.818)。药敏试验结果显示,GBS对青霉素、头孢噻肟、利奈唑胺和万古霉素等的体外敏感率均为100.00%。结论孕妇年龄、孕次、孕前检查、流产史、妊娠期阴道炎及妊娠期糖尿病与GBS阳性有关,基于GNB算法构建的联合预测模型是预测围生期孕妇GBS阳性的良好模型。Objective To explore the clinical characteristics associated with group B Streptococcus(GBS)infection in perinatal pregnant women so as to construct a prediction model for GBS infection,providing more reference for improving pregnancy outcomes.Methods A retrospective study was conducted on 1,538 perinatal pregnant women from the obstetrics outpatient and inpatient departments of the First,Second,and Third People's Hospitals of Xiaoshan District,Hangzhou,from January to July 2021.The participants were divided into GBS-positive and GBS-negative groups based on GBS culture results,and their clinical data were collected.Seven artificial intelligence algorithms[logistic regression,gaussian naive bayes(GNB),multi-layer perceptron,support vector machine,random forest,light gradient boosting machine,and decision tree]were combined with clinical indicators to develop GBS prediction models.The dataset was split into training and validation sets at an 8∶2 ratio.The model performance was evaluated using the area under the receiver operating characteristic curve(AUC),sensitivity,specificity,and accuracy.Additionally,antimicrobial susceptibility testing was performed on the GBS strains.Results Among the 1538 participants,98 were GBS positive,with a positive rate of 6.37%.Univariate analysis showed that maternal age,number of pregnancies,pre-pregnancy check-up,history of miscarriage,gestational vaginitis,and gestational diabetes were associated with GBS infection(all P<0.05).The GBS infection prediction model combining the six clinical features under the"GNB"algorithm demonstrated the best diagnostic performance(AUC=0.800,sensitivity=0.675,specificity=0.818).Antimicrobial susceptibility testing showed that GBS had a 100.00%in vitro sensitivity rate to penicillin,cefotaxime,linezolid,and vancomycin.Conclusion Maternal age,number of pregnancies,prepregnancy check-up,history of miscarriage,gestational vaginitis,and gestational diabetes are associated with GBS infection.The combined prediction model based on the"GNB"algorithm is

关 键 词：围生期孕妇 B族链球菌 人工智能 预测模型 

分 类 号：R71[医药卫生—妇产科学]