Unsupervised Classification in Uterine Electromyography Signal: Toward The Detection of Preterm Birth

Abstract

The purpose of this study is to classify the uterine contractions in the electromyography (EMG) signal. As the frequency content of the contraction changes from one woman to another and during the pregnancy, wavelet decomposition is used to extract the parameters of each contraction, and an unsupervised statistical classification method based on competitive artificial neural network is then used to classify events. A principal component analysis projection is then used to evidence the groups resulting from this classification. Results show that uterine contractions may be classified into independent groups according to their frequency content and so according to the pregnancy terms. This classification will be used to detect the preterm birth.