基于人工神经网络的剖宫产术后再次妊娠阴道分娩的预测研究
|
摘要
目的建立基于人工神经网络(ANN)的剖宫产术后再次妊娠阴道分娩(VBAC)预测模型,并比较该模型与Logistic回归模型的预测能力。方法选取2015年1月-2017年12月在青岛市城阳区人民医院住院并剖宫产后阴道试产(TOLAC)的孕妇为研究对象,根据是否成功阴道分娩,将产妇分为VBAC组(354例)和重复剖宫产组(RCS,76例)。采用SPSS 24. 0软件包ANN模块中的多层感知器(MLP),构建预测VBAC的最佳ANN模型。输入参数设为:分娩年龄、孕期体质量增加量、孕周、孕次、第一产程时间、有无剖宫产指征、是否人工破膜、有无脐带绕颈、是否患有妊娠期糖尿病(GDM)、Bishop评分、胎儿双顶径和股骨长度,共12个变量。输出参数设为:分娩方式。并采用逐步向后二元Logistic回归对以上指标进行分析,构建预测模型。以实际分娩方式作为金标准,比较两种预测模型的曲线下面积(AUC)是否存在差异。结果产妇年龄越小、孕期体质量增加较少、无剖宫产指征、自然破膜以及Bishop评分越高的妇女VBAC的可能性越高,差异有统计学意义(P<0. 05)。ANN模型和Logistic回归模型AUC分别为0. 850 (95%CI:0. 804～0. 897)和0. 810 (95%CI:0. 756～0. 863),表明这两种预测模型都具有较高的预测能力。虽然ANN预测模型与Logistic回归预测模型AUC差异无统计学意义,但是ANN预测模型的预测能力稍高于Logistic回归模型。结论运用ANN构建VBAC预测模型具有较高的预测能力,不仅可为VBAC风险预测提供基础理论和方法支持,也有利于早期筛选适合TOLAC的孕妇以及早期发现剖宫产术后再次妊娠妇女子宫破裂的倾向,具有广阔的应用前景。
Objective To establish prediction model of vaginal delivery after cesarean section( VBAC) based on artificial neural network( ANN),compare the predictive ability of VBAC model and Logistic regression model. Methods The hospitalized pregnant women receiving trial of labor after cesarean delivery( TOLAC) in People's Hospital of Chengyang District from January 2015 to December 2017 were selected and divided into ANN group( 354 cases) and repeated cesarean section group( RCS group,76 cases) according to successful vaginal delivery or not. Multilayer perceptron( MLP) of SPSS 24. 0 software package ANN module was used to construct optimal ANN model of predicting VBAC( P<0. 05). The input parameters included maternal age,weight gain during pregnancy,gestational week,gravidity,the time of the first stage of labor,indication of cesarean section,artificial rupture of fetal membrane,cord around neck,gestational diabetes mellitus( GDM),Bishop score,fetal biparietal diameter,and femur length; the output parameter was delivery mode. Stepwise backward binary Logistic analysis was used to analyze the above-mentioned indexes and construct prediction model. Taking the actual delivery mode as the gold standard,the differences of the area under ROC curve( AUC) were compared between the two prediction models. Results The young pregnant women,the pregnant women with less weight gain during pregnancy,without indications of cesarean section,with natural rupture of fetal membrane,and high Bishop score had high probablity of VBAC. The values of AUC of ANN model and Logistic regressoin model were 0. 850( 95% CI: 0. 804-0. 897) and 0. 810( 95% CI: 0. 756-0. 863),respectively,which showed that the prediction abilities of the two models are high. Although there was no significant difference in AUC between ANN prediction model and Logistic regression prediction model,the prediction ability of ANN prediction model was slightly higher than that of Logistic regression model. Conclusion The prediction ability of ANN prediction model is high,which not only can provide basic theory and method support for VBAC risk prediction,but also be helpful for early screening of pregnant women suitable for TOLAC and early detection of uterine rupture in women after cesarean section. This model has broad application prospect.
Objective To establish prediction model of vaginal delivery after cesarean section( VBAC) based on artificial neural network( ANN),compare the predictive ability of VBAC model and Logistic regression model. Methods The hospitalized pregnant women receiving trial of labor after cesarean delivery( TOLAC) in People's Hospital of Chengyang District from January 2015 to December 2017 were selected and divided into ANN group( 354 cases) and repeated cesarean section group( RCS group,76 cases) according to successful vaginal delivery or not. Multilayer perceptron( MLP) of SPSS 24. 0 software package ANN module was used to construct optimal ANN model of predicting VBAC( P<0. 05). The input parameters included maternal age,weight gain during pregnancy,gestational week,gravidity,the time of the first stage of labor,indication of cesarean section,artificial rupture of fetal membrane,cord around neck,gestational diabetes mellitus( GDM),Bishop score,fetal biparietal diameter,and femur length; the output parameter was delivery mode. Stepwise backward binary Logistic analysis was used to analyze the above-mentioned indexes and construct prediction model. Taking the actual delivery mode as the gold standard,the differences of the area under ROC curve( AUC) were compared between the two prediction models. Results The young pregnant women,the pregnant women with less weight gain during pregnancy,without indications of cesarean section,with natural rupture of fetal membrane,and high Bishop score had high probablity of VBAC. The values of AUC of ANN model and Logistic regressoin model were 0. 850( 95% CI: 0. 804-0. 897) and 0. 810( 95% CI: 0. 756-0. 863),respectively,which showed that the prediction abilities of the two models are high. Although there was no significant difference in AUC between ANN prediction model and Logistic regression prediction model,the prediction ability of ANN prediction model was slightly higher than that of Logistic regression model. Conclusion The prediction ability of ANN prediction model is high,which not only can provide basic theory and method support for VBAC risk prediction,but also be helpful for early screening of pregnant women suitable for TOLAC and early detection of uterine rupture in women after cesarean section. This model has broad application prospect.
引文
[1]Lumbiganon P,Laopaiboon M,Gulmezoglu AM,et al.Method of delivery and pregnancy outcomes in Asia:the WHO global survey on maternal and perinatal health 2007-08[J].Lancet,2010,375(9713):490-499.
[2]Naji O,Wynants L,Smith A,et al.Predicting successful vaginal birth after Cesarean section using a model based on Cesarean scar features examined by transvaginal sonography[J].Ultrasound Obstet Gynecol,2013,41(6):672-678.
[3]Metz TD,Stoddard GJ,Henry E,et al.Simple,validated vaginal birth after cesarean delivery prediction model for use at the time of admission[J].Obstet Gynecol,2013,122(3):571-578.
[4]Grobman WA,Lai Y,Landon MB,et al.Does information available at admission for delivery improve prediction of vaginal birth after cesarean?[J].Am J Perinatol,2009,26(10):693-701.
[5]张会敏,叶明全,罗永钱,等.基于RBF神经网络的老年痴呆症智能诊断研究[J].中国数字医学,2015,10(6):38-41.
[6]王晓华,陈卉,马大庆,等.人工神经网络在孤立性肺结节CT诊断研究中的应用[J].中华放射学杂志,2006,40(4):377-382.
[7]黄仕鑫,罗佳婧,罗亚玲,等.基于BP神经网络模型鉴别2型糖尿病肾病的认知图研究[J].中华内分泌代谢杂志,2017,33(11):943-949.
[8]中华医学会妇产科学分会产科学组.剖宫产术后再次妊娠阴道分娩管理的专家共识(2016)[J].中华妇产科杂志,2016,51(8):561-564.
[9]伍绍文,卢颖州,王珊珊,等.剖宫产术后再次妊娠阴道分娩的相关影响因素分析[J].中华妇产科杂志,2016,51(8):576-580.
[10]李博雅,杨慧霞.剖宫产后再次妊娠热点问题[J].医学与哲学,2017,38(20):18-21.
[11]Landon MB,Leindecker S,Spong CY,et al.The MFMU Cesarean Registry:factors affecting the success of trial of labor after previous cesarean delivery[J].Am J Obstet Gynecol,2005,193(3 Pt2):1016-1023.
[12]郭晋.神经网络模型在预测急性心肌梗死中的应用及模型预测能力的比较研究[D].北京:北京协和医学院,2013.
[13]吴拥军,吴逸明,屈凌波,等.人工神经网络技术在肺癌诊断中的应用研究[J].中华微生物学和免疫学杂志,2003,23(8):646-649.
[14]田敬霞,张景祥.人工神经网络预测分娩方式的研究[J].医学信息(中旬刊),2011,6(8):3921-3922.
[15]李昆,柴玉梅,赵红领,等.基于深度神经网络的胎儿体重预测[J].计算机科学,2016,28(S2):73-76.
[16]吴妍,姚蕾,盛文丽,等.基于人工神经网络的卵巢早衰预测模型研究[J].中国全科医学,2017,20(27):3410-3415.
[2]Naji O,Wynants L,Smith A,et al.Predicting successful vaginal birth after Cesarean section using a model based on Cesarean scar features examined by transvaginal sonography[J].Ultrasound Obstet Gynecol,2013,41(6):672-678.
[3]Metz TD,Stoddard GJ,Henry E,et al.Simple,validated vaginal birth after cesarean delivery prediction model for use at the time of admission[J].Obstet Gynecol,2013,122(3):571-578.
[4]Grobman WA,Lai Y,Landon MB,et al.Does information available at admission for delivery improve prediction of vaginal birth after cesarean?[J].Am J Perinatol,2009,26(10):693-701.
[5]张会敏,叶明全,罗永钱,等.基于RBF神经网络的老年痴呆症智能诊断研究[J].中国数字医学,2015,10(6):38-41.
[6]王晓华,陈卉,马大庆,等.人工神经网络在孤立性肺结节CT诊断研究中的应用[J].中华放射学杂志,2006,40(4):377-382.
[7]黄仕鑫,罗佳婧,罗亚玲,等.基于BP神经网络模型鉴别2型糖尿病肾病的认知图研究[J].中华内分泌代谢杂志,2017,33(11):943-949.
[8]中华医学会妇产科学分会产科学组.剖宫产术后再次妊娠阴道分娩管理的专家共识(2016)[J].中华妇产科杂志,2016,51(8):561-564.
[9]伍绍文,卢颖州,王珊珊,等.剖宫产术后再次妊娠阴道分娩的相关影响因素分析[J].中华妇产科杂志,2016,51(8):576-580.
[10]李博雅,杨慧霞.剖宫产后再次妊娠热点问题[J].医学与哲学,2017,38(20):18-21.
[11]Landon MB,Leindecker S,Spong CY,et al.The MFMU Cesarean Registry:factors affecting the success of trial of labor after previous cesarean delivery[J].Am J Obstet Gynecol,2005,193(3 Pt2):1016-1023.
[12]郭晋.神经网络模型在预测急性心肌梗死中的应用及模型预测能力的比较研究[D].北京:北京协和医学院,2013.
[13]吴拥军,吴逸明,屈凌波,等.人工神经网络技术在肺癌诊断中的应用研究[J].中华微生物学和免疫学杂志,2003,23(8):646-649.
[14]田敬霞,张景祥.人工神经网络预测分娩方式的研究[J].医学信息(中旬刊),2011,6(8):3921-3922.
[15]李昆,柴玉梅,赵红领,等.基于深度神经网络的胎儿体重预测[J].计算机科学,2016,28(S2):73-76.
[16]吴妍,姚蕾,盛文丽,等.基于人工神经网络的卵巢早衰预测模型研究[J].中国全科医学,2017,20(27):3410-3415.