Classification of Uterine Electromyogram Signal Using Auto-Bicoherence Spectrum and Neural Network Classifier for Detecting Preterm Birth

Abstract

The prediction of premature deliveries is a major problem.  A lot of research uses different tools and physiological parameters to predict these types of births and to take care of the mother and the premature baby. Uterine contraction signals or uterine electromyogram (EMG) contain useful information in the prediction of premature birth. In this work we used the spectrum of the uterine EMG signals' auto-bicoherence. This analysis is used to extract the parameters that characterize a preterm delivery from a full-term delivery. The parameters are then used by a classifier.