Vinsamlegast notið þetta auðkenni þegar þið vitnið til verksins eða tengið í það: http://hdl.handle.net/1946/22974
The uterine EMG -called Electrohysterogramme (EHG)- temporal, frequency, and time-frequency characteristics have been used for a long time for the prediction of preterm labor. However, the investigation of its propagation is rare. All the results of the previous studies did not show a satisfactory potential for clinical application. The objective of this thesis is the analysis of the propagation as well as of the nonlinear characteristics of EHG signals during pregnancy and labor for clinical application. A monovariate analysis was done to investigate the nonlinearity and the sensibility of methods to different characteristics of the signals. A bivariate analysis was then done for the investigation of the propagation of EHG by measuring the coupling between channels, as well as the direction of coupling, which is an innovative part of our thesis. In this analysis we propose a new approach to improve the coupling and direction estimation methods. Another innovation of this thesis is the implementation of a tool for EHG source localization to investigate the dynamic of the uterus at the source level, not at electrodes level as previously done. Results show that nonlinear methods are more able to classify pregnancy and labor contractions than linear ones, and that time reversibility method is the least sensitive to sampling frequency and frequency content of the signal. Results also indicate an increase in coupling and a concentration of coupling direction toward the cervix when going from pregnancy to labor. We also proposed to respect the nonstationarity of EHG signal and to recover the effect of variable fat filtering along pregnancy, by segmenting and filtering the EHG in its FWL component. This filtering-windowing approach permits to improve the performances of connectivity methods. Finally, the intensity of localized sources and their number is higher in labor than in pregnancy contractions. The identified sources are more active and more propagated in labor whereas in pregnancy they remain weak and local. An improvement in the electrode matrix of the rat experimental protocol has also been done by developing a suction electrode. This protocol can then be used for the validation of our methods and of the electrophysiological model.