Research On Several Machine Learning Algorithms And Application In Fetal Heart Monitoring

Fetal distress is a syndrome that occurs in the post-natal or late-pregnancy period,which can lead to lack of oxygen in the fetus,severe distress will affect fetal growth and even cause fetal death,timely and accurate classification and appropriate medical measures are the main means to reduce the risk of childbirth.With the wide spread of artificial intelligence in the medical industry,medical data is becoming more and more standardized and comprehensive under extensive attention,and machine learning algorithm can learn the inherent features from a large number of data and provide classification or prediction results,which is of great significance for saving medical resources and improving the efficiency and accuracy of diagnosis.In this paper,several machine learning algorithms are used to study the clinical data of fetal distress,so as to learn the pathological features that can be recognized by the computer from the data,establish a model and accurately classify the pathological features so as to achieve the goal of machine intelligent diagnosis.This paper mainly completes the following work:1.The relevant theoretical knowledge of fetal distress is introduced,including the period of fetal distress,the international classification criteria for fetal distress,and the causes of fetal distress.2.Three classification tasks are implemented for discrete data of fetal heart monitoring in UCI.The Stacking ensemble learning model was compared with the BP neural network,SVM support vector machine,RF random forest,AdaBoost algorithms,and it was found that the Stacking model had the best effect in the classification recognition of fetal heart monitoring,and its classification accuracy,accuracy,recall rate,F1 comprehensive indexes reached 95.82%,97.01%,97.98%,97.49%.3.Analyzed the electronic fetal heart monitoring time series data in PhysioNet,according to the characteristics of the data and the objectives of this paper,GADF was used to convert the time series data of fetal heart rate(FHR)and uterine contraction(UC)into image data and merge them.Then using six layers of CNN model and VGG16 model to predict fetal distress important indicators: Apgar score,pH value of umbilical cord arterial blood.The comparison results showed that the evaluation effect of VGG16 model was better,with the mean square error reaching 0.05779076,0.03113010.