Analyse du Rythme Cardiaque Foetal par des Methodes de Detection de Ruptures

Abstract

The object is to elaborate a method of automatic quantitative and real time analysis of the Fetal Heart Rate, using concopt of signal processing and data analysis. There are three parameters to extract : the basic mean rate, the accelerations-decelerations and the periods of low and high variations. The accelerations-decelerations, in regard to the basic rate, are found with the Hinkley algorithm by detecting the rupture in mean, and then declared valid, according to global criterions. The variations are studied by modeling, a high frequency component of the signal, by mean of linear prediction. The eventual ruptures of the model’s coefficients are searched, by using an algorithm developped by M. Basseville and A. Benveniste. So the component is segmented in episodes of stationarity. These episodes are gathered according to the standard deviation and the model’s coefficients, with methods of automatic clustering. The standard deviation is the more discriminant variable. The results thus obtained and those from the visual analysis of the curve are quite consistent. Further more, additional informations about the spectrum of each episode are provided.