Abdominal Electrodes Analysis by Statistical Processing for Fetal Electrocardiogram Extraction

Obstetricians were asking the engineering support to study more extensively any technical possibility to electronically get some useful information from the whole PQRST complex of the fetal electrocardiogram (fECG) in order to identify eventual sign of fetal distress (FD). The latter should indeed be found as a more reliable information for the clinician with the potential benefit of increasing the sensibility as well as the specificity of the diagnosis of FD. A way to achieve this goal is Blind Source Separation. In this case, the extraction of an estimation of the fetal PQRST complex can be solved using Independent Component Analysis on signals recorded through electrodes located on the pregnant woman abdomen. Today, neither theoretical nor simulation considerations were investigated to determine an optimal number and location of these electrodes, despite possible important consequences on the extraction performances. We propose here a method to identify the location of electrodes (depending of the position of the fetus) that drive electrical components due to the fetus by analyzing the (joint and marginal) density functions of the recorded signals. This result allows to evaluate the ‘interesting’ sensors, therefore allowing electrode selection when many sensors are involved in the measure. We show several simulation results on artificial signals. Key Words Fetal electrocardiogram analysis, non-invasive measurement, mutual information, blind source separation.