Analysis and Classification of Three Trimesters During Normal Pregnancy Using Bispectrum

Abstract

This study presents the analysis and classification of maternal status during healthy pregnancy and postpartum using bispectral features of heart rate variability (HRV). Bispectral analysis is applied to HRV to extract a total of 11 features out of which only 6 significantly different feature values are used to classify the subjects into three trimesters during pregnancy and postpartum. In particular, bispectrum patterns indicate the presence of higher phase coupling during postpartum as compared to the pregnancy group. The decreased phase coupling is observed as pregnancy progresses that refer to the decreased nonlinear interactions during pregnancy. This infers that the pregnancy is characterized by decreased HRV due to a reduced vagal tone instead of increased sympathetic tone. The six selected features are used as input to the k-nearest neighbour (KNN), Gaussian mixture model (GMM) and probabilistic neural network (PNN) classifiers for data classification. The performance of the classifiers are measured based on the accuracy of classification. The classification rate obtained for KNN, GMM and PNN classifiers are 85.94%, 85.94% and 89.74% respectively. We conclude that the bispectral features with PNN classifier is better suitable for the classification of the three trimesters during pregnancy and postpartum.