Classification of Nonlinear Features of Uterine Electromyogram Signal Towards the Prediction of Preterm Birth

Abstract

Early detection of preterm labor is important to avoid neonatal death and mortality. Uterine electromyography (UEMG) or electrohysterography is a non-invasive method of extracting electrical activity signal from the abdominal part during pregnancy, which helps in early detection. This signal can be used to classify term and preterm labors. Herein, the performances of four classifiers have been evaluated using seven nonlinear features extracted from UEMG signals. They were then compared with four features analyzed from different literature. The results show that with the Elman neural network classifier, the bi-spectrum feature, which has phase information, outperforms other features with 99.8875% accuracy, 100% sensitivity, and 99.77% specificity.