Channel combination selection for EHG bivariate analysis

Abstract

The EHG signals are recorded using a multichannel system positioned on the woman's abdomen for the simultaneous recording of 16 channels of EHG. Several studies calculated the features related to EHG propagation by studying the coupling between all possible channels (bivariate analysis). Using all the possible features extracted to characterize propagation, from all possible combinations of channels, lead to a very large dimension of search and to a complex classification. Therefore, the aim of this paper is the selection of the most relevant channel combinations (using Fscore method), that provide the most useful information to discriminate between pregnancy and labor classes. This channel combination selection step is then followed by a feature selection method named genetic algorithm that is used to select the best features (from the propagation features used in this study) from the selected channel combinations. Additionally, we applied these selection steps on bipolar and monopolar EHG signals in order to see which is the best to use for the bivariate analysis.