Predicting Preterm Delivery Based on Wavelet Packet Analysis of a Single Electrohysterography Channel

Abstract

Predicting preterm delivery is of great importance as it can help to avoid certain medical complications and increase the chance of survival for the newborn. This work presents a new method to predict the possibility of preterm delivery. A total of 162 records were used in this study, 143 records ended with fullterm delivery and 19 records ended with preterm delivery. The prediction algorithm is based on the analysis of single Electrohysterography recording (EHG) using wavelet packets. The EHG signal was decomposed into three-level wavelet packets transform and features were calculated using the sample entropy of each wavelet packet. The obtained results showed an improvement in comparison to other works in the literature. For the test data set, accuracy was found to be 95%, sensitivity 100%, and specificity 75%, with a kappa coefficient of 83%.