Uterine EMG Analysis: Time-Frequency Based Techniques for Preterm Birth Detection - CNRS - Centre national de la recherche scientifique Accéder directement au contenu
Chapitre D'ouvrage Année : 2009

Uterine EMG Analysis: Time-Frequency Based Techniques for Preterm Birth Detection

Résumé

The global aim of this work is to detect preterm deliveries using the uterine electromyography signal. For this purpose, two steps are required: the first step aims to detect all events in this signal and to identify these events by allocating them to physiological classes: contractions, foetus motions, Alvarez or Long Duration Low Frequency (LDBF) waves. The second step consists of the identification of contractions between normal contractions and preterm birth contractions. Detection and identification of events are based on the use of Wavelet Packet Transform (WPT) to select the best basis for both detection and classification goals. This is achieved by selecting the only WPs of the decomposition tree that are able to highlight changes in the recordings using a training set of signals. The selection criterion is based on the Kullback Leibler distance and the Generalised Gaussian Distribution (GGD). The detection algorithm “Dynamic Cumulative Sum” (DCS) is applied on selected packets. This combination of DCS and WP decomposition has been shown to be very efficient for the detection of both frequency and energy changes. After obtaining the uterine events, classification consists of identifying the detected events. The ratio between the intra-class variance and the total variance (sum of the inter-class and intra-class variances) is used as a criterion for the best basis of classification. In most cases, more than 85% of events are well detected and classified whatever the term of gestation. Finally, the classification of contractions is achieved using wavelet networks. This network is learned using normal contractions and preterm births contractions. As final results, uterine contractions are well classified whatever the term of gestation and this global system (detection, classification, identification and diagnostic) provides good results concerning the preterm births.

Dates et versions

hal-00456005 , version 1 (11-02-2010)

Identifiants

Citer

Mohamad Khalil, Marwa Chendeb, Mohamad Diab, Catherine Marque, Jacques Duchene. Uterine EMG Analysis: Time-Frequency Based Techniques for Preterm Birth Detection. Advanced Biosignal Processing, Advanced Biosignal Processing » A. Naït-Ali Ed., Springer; Springer Berlin Heidelberg, pp.245-266, 2009, ⟨10.1007/978-3-540-89506-0_12⟩. ⟨hal-00456005⟩
133 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More