Classification of Term and Preterm Birth Data from Elektrohisterogram (EHG) Data by Empirical Wavelet Transform Based Machine Learning Methods

Accurate prediction of preterm birth can significantly reduce birth complications for both mother and baby. This situation increases the need for an effective technique in early diagnosis. Therefore, machine learning methods and techniques used on Electrohysterogram (EHG) data are increasing day by day. The aim of this study is to evaluate the effectiveness of the Empirical Wavelet Transform (EWT) approach on EHG data and to propose an algorithm for estimating preterm birth using single EHG signal. The data used in the study were taken from Physionet's Term-Preterm Electrohysterogram Database (TPEHGDB) and scored in one-minute windows. The feature matrix was obtained by calculating the sample entropy value from each of the discretized EHG modes obtained as a result of this method, which was used for the first time on EHG data, and the average energy value from the signal obtained by recombining the modes. The obtained features were applied to Random Forest (RF), Support Vector Machine (SVM), Long Short-Term Memory (LSTM) algorithms to predict preterm birth. Among the classifier algorithms, the RF algorithm achieved the best result with a success rate of 98,20%.