The Application of Machine Learning Models in Fetal State Auto-Classification Based on Cardiotocograms

2019 2nd International Conference of Green Buildings and Environmental Management

生态环境科学

Xue, Gongao^1

Sino-German College of Technology, East China University of Science and Technology, Shanghai

200237, China^1

关键词: Biology and medicine;    Cardiotocography (CTG);    Classification models;    Machine learning models;    Physical conditions;    State classification;    Training and testing;    Uterine contractions;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/310/5/052007/pdf
DOI  :  10.1088/1755-1315/310/5/052007

学科分类：环境科学（综合）

来源: IOP

【 摘 要 】

Cardiotocography (CTG) is widely used by obstetricians in accessing the physical condition of a fetus during pregnancy, for it provides obstetricians with the data regarding the fetal heartbeat and the uterine contractions which helps determine whether the fetus is pathologic or not. Traditionally obstetricians analyze data from CTG artificially, which is both time consuming and lack of reliability. For this reason, developing a fetal state auto-classification model is necessary, for it can not only reduce the time for diagnosing but also save medical resources. With machine learning developing rapidly nowadays, it has been widely applied in areas like biology and medicine to solve various problems. In the condition of fetal state classification, we apply neural network and random forest to analyze the cardiotocographic data from the UCI Repository. Since there is high imbalance in our data, method of weighing has also been applied to optimize our model. Random forest outperforms neural network in terms of accuracy in classifying types of fetuses, which achieves 88.84% and 91.85% accuracy on the training and testing set respectively.