BP神经网络预测妊娠期糖尿病胎儿体重的研究.docx

BP神经网络预测妊娠期糖尿病胎儿体重的研究
 作者:宋鹤兰,张东枚,李丽霞,李筠,曾慧韵,丁淑瑾,赵曼丹,骆婕 
【摘要】 目的探讨BP神经网络预测妊娠期糖尿病胎儿出生体重的价值。方法将306例足月、单胎、无妊娠其它合并症及并发症的GDM孕妇随机分为训练组和验证组。训练组分别选取不同参数构建3个神经网络:(1)孕妇参数法:包括孕妇体重指数、腹围、宫高、孕期增加体重、空腹血糖、餐后2h血糖、糖化血红蛋白等7项参数作为输入节点;(2)胎儿参数法:用胎儿的双③顶径、股骨长度、头围、腹嘶围、腹径、股骨皮下脂肪厚葺度、胎儿腹壁脂肪层厚度等7项参数作为输入节点;(蜡3)联合参数法:将孕妇及号胎儿的参数作为输入节点。溢神经网络构建完成后以10踝6例验证组来分别测试3种穿网络法的误差率和符合率。诗结果联合参数法准确率最高单为%,胎儿参数法为%,孕绣妇参数法为%。结论BP神稣经网络预测胎儿体重有很好哄的应用前景。选取合适的孕妇及胎儿参数建立网络可提彝高预测的准确性。
【关键铑词】 妊娠期糖尿病;胎儿薹体重;预测;BP神经网络改
Abstract:Ob音jective Toin氨vestigatethe獐
workin阙predictingfe膳talbirthwEig咔htinpatients⒆withgestatio媛naldiabetesmellitus(GDM)termpregnancy,sin差glegestation幺,plicat躐ionsofGDM,we咚rerandomlydividedintotra葑ininggroup(2茄00cases,including106male枷fetusesand94埏femalefetuses)andtestgro原up(106cases,including56m仙alefetusesanΨd50femalefet跨uses).Traini镣nggroupweres歪electedtobui碍ldthreediffe玷rentneuralne﹃tworkswithdi魏fferentparam锸eters,(l)Pre遗gnantwomenparametermetho叙d:includingb馁odymassindex翅(BMI),abdomi街nalcircumfer焉ence,fundalh牺EIght,pregna啬ncyweightgai沤n,fastingblo妹odsugar(FBS)蝶,postprandia砣lbloodsugar(酒PBS),glycosylatedhemoglo纡bin(GHbA1c),祓thesesevenpa埒rameterswereusedasinputn素odes.(2)Feta镓lparameterme铃thod:includi诈ngfetalbipar野ietaldiamete苡r(BPD),femur噼length(FL),headcircumfer崮ence(HC),abd貉ominalcircum露ference(AC),abdominaldia腙meter(AD),fe五moralthighso噼fttissuethic敲kness(FTSTT)踵,andfetalabd怡ominalwallfa祝tlayer(FFL),蜕thesesevenpaдrameterswereusedasinputn帮odes.(3)Join微tparametermethod:shmentofneur潼
works,thedataof106c炉asesoftestgr褐oupwasusedto晋verifiedpred踟ictionerrorr氚ateandpredictioncoincide畿ncerateofthe晴threeneuraln目etworksinpre向dictingfetalbirthweight.
Results urate,turacyra眺teofthismeth副odwasupto%,urac您yratesoffeta猪lparametersm缩ethodandmoth擤erparametermethodwere%an菖d%全workpredicti殛onoffetalbir╇