Uterine EMG processing : dynamic detection associated with multiscale classification of events
Résumé
Towards the goal of detecting preterm birth by characterizing the events in the uterine electromyogram (EMG), we propose a new approach for detection and classification of events in this signal. Detection is based on the Dynamic Cumulative Sum (DCS) of the local generalized likelihood ratio associated with a multiscale decomposition using wavelet transform. An unsupervised classification based on the comparison between variance-covariance matrices computed from selected scales has been implemented after detection. Finally a class identification based on a neural network is used. This algorithm of detection-classification-labelling gives satisfactory results on uterine EMG: in most cases more than 80% of events are well-detected and classified whatever the term of gestation.