A nonlinear parameterization of multivariate electrohysterographical signals

Electrohysterography is a technique which measures a bioelectrical activity of a uterus. This paper presents an application of a nonlinear parameterization of multivariate electrohysterographical signals for a uterine activity assessment to improve unsatisfactory a labor prediction accuracy by methods published in the literature. A multivariate sample entropy used for differentiated 4-channel electrohysterographical signals, general Spearman׳s correlation and a combined index being the sum of them, were tested. These nonlinear measures use joint information contained in a multivariate signal. The results confirm that the combined index provides the best assessment of uterine contractions: 87% sensitivity and 50% specificity of labor prediction in the studied data. These results should be verified in a prospective study.