Research Of Contraction Identification Algorithm During Fetal Monitoring

Antenatal cardiotocography(CTG),as a most common and most useful monitoring method of assessing the status of fetus,is used widely to reduce the percent of miscarriage,preterm rate and perinatal mortality effectively.Uterine contraction(UC)is a kernel parameter of electronic fetal monitoring in perinatal,which is related to the status of the fetus.Frequent and strong UC can lead to delivery,causing the fetus mortality or permanent damage to fetus due to fetal hypoxia under the stress from high density contraction.Therefore,it's clinically significant for the prevention of preterm labor to analysis UC signals to extract physiological information related to fetus health.The paper mainly studied the uterine contraction processing and intelligent identification algorithm during fetal monitoring.First,the basic features of UC signal were studied and analyzed,and an adaptive filtering method based on signal's quality classification was presented.Then,the test results of baseline were analyzed.Secondly,the UC identification algorithm based on double finite state machines(FSM)was studied.The construction and relationship of the two FSMs were introduced.After extracting the features of the signals from the database,a best decision tree was constructed through the classification algorithm based on decision tree to realize the intelligent classification of UC intensity.The main works of this paper are mainly as followed:(1)The basic knowledge of UC signal and the methods of UC collection were introduced.And the main ways of UC measurement were presented.Then the assessment indexes of algorithm were investigated,followed by the introduction of two databases used in paper.These provided the basic theoretical support for its processing and identification.(2)Two classification standards for UC signals were present.One is based on spectrum estimation of autoregressive(AR)model,and another one is modified STD(Standard Deviation).The sampled UC data were classified based on the two classification criteria.A suitable filtering method was selected after signal classification adaptively.Finally,a statistical central principle of the layer distribution was presented to realize the detection and modification of the UC signals.(3)Uterine contraction algorithm based on double FSMs was introduced.Basic of the FSM was presented first.The construction of two FSMs was executed followed by the description of state changes of FSMs and the relationship between them.Then,how to recognize a UC by double FSMs and how to count UC were introduced.Finally,the proposed algorithm was performed on the two databases to test its performance.(4)The classification algorithm based on decision tree was proposed for the classification of UC strength.The features of UC signals were extracted at first.Then the change rule between the parameter k of K-fold cross validation and the performance of the classification algorithm was investigated.The impact varied features under different optimal selecting attribute method made on to the performance of classification was well presented.Finally,the proposed classification method was compared with other work.This paper finished the accurate identification UC occurrence and intelligent classification of UC strength,and provided a theoretical basis and technical support for the engineering application of UC identification and classification algorithm.