A New Efficient Algorithm for Prediction of Preterm Labor

Electrohysterogram (EHG) signal represents electrical activity of uterine collected from abdominal surface of pregnant women. It has been proven that EHG analysis could be a suitable way to predict preterm labor and consequently to prevent it. The aim of this paper is to present an efficient low computational complexity algorithm to detect preterm labor using EHG signals. To this purpose, Empirical Mode Decomposition (EMD) has been applied for features extraction. Root mean square (RMS) of first two intrinsic mode function (IMF) of decomposed EHG signals were used as features and classification was performed by Support Vector Machine (SVM). In this experiment, the database consisted of 262 term delivery subjects (duration of pregnancy >= 37 weeks) and 38 preterm delivery subjects (duration of pregnancy < 37 weeks). Each record contained 3 channels, therefore, various configurations of features and channels were applied. Among those configurations, classification based on RMS of the second IMF from channel one achieved best results with accuracy = 99.56%, sensitivity = 98.95% and specificity = 99.30%.