Study On The Method Of Predicting Prenatal Implantation Of Placenta

Placenta accreta(Placenta accreta,PA)refers to the placenta villi invaded into the myometrium and rooted in the uterine wall,the implantation part can not be normal self-peeling.Moreover,the artificial stripping will damage the myometrium and may lead to massive bleeding,uterine perforation,shock,secondary infection,even death,and some patients were forced to perform hysterectomy due to severe postpartum hemorrhage.In the medical field,the studies of placenta implantation always get widespread attention from medical scientists at home and abroad.However,the traditional researches focused on the effectiveness and safety of surgery.Meanwhile,most researches regarding to the prevention and prenatal diagnosis of placental implant disease are qualification,and still in the exploratory stage.In order to achieve the prenatal intelligent prediction of PA,this subject established a corresponding data model aimed at the disease condition of PA which was based on the current medical data.Above all,according to the characteristics of the clinical data set of PA,we performed the feature extraction regarding to the related factors of the incidence of PA.After that,we built the hidden Markov model of PA based on this feature set,and the prediction accuracy is improved by optimizing the feature weights.Specifically,this paper mainly includes the following three parts:(1)A PA feature extraction algorithm based on combining gray comprehensive correlation and entropy methoh.As the disease condition of PA is complex and the factors which caused PA are numerous,it can not judge whether the pregnant women is sick or not and the degree of disease condition from the prenatal diagnostic data.Therefore,according to the multiple related features and difficulties in determination of PA,this study proposed a method which combined gray comprehensive correlation and entropy methoh to analyze the prenatal diagnostic data,extracted the set of related factors which induced the occurrence of PA.Experiments show that the algorithm selected in this paper can improve the accuracy of relevance factor extraction under the premise of reducing the interference factor caused by human compared with Relief algorithm and pure gray comprehensive correlation degree algorithm.(2)PA-based hidden Markov model.According to the characteristics of the disease condition of PA,it needed to extract the set of related factors which including multiple features to evaluated the disease condition of pregnant women.Because the observation state is single in the traditional Hidden Markov Model,it is not suitable for predicting PA.Thus,this research built a multi-state hidden Markov model PA-HMM which finally obtained the morbidity rate of PA by the iterative computation of models,and used it to evaluate the disease condition of pregnant women.Through the experiment contrasts,PA-HMM is superior to traditional methods such as SVM and NN in accuracy rate.(3)Model optimization and the design and achievement of predicted software.In the above study,as the weight ratio of different features are identical in the Hidden Markov Model which constructed by the related factors of PA,we needed optimize the model of PA-HMM.This study assigned the corresponding weight coefficients to this model by the method which combined the PCA and comentropy.Furthermore,based on the data of clinical medical examination,the function modules of PA prenatal prediction software are designed,and finally completed the system software.