Estimation of Foetal Contribution to ECG Recordings Using Oblique Projection Technique Exploiting Cyclostationary Properties of the Heartbeat Signals

In this paper, the authors propose to estimate the contribution of the electrical activity of the foetal heart to each component of cutaneous electrocardiogram (ECG) recordings using an oblique projection technique. To date, related research work reported in the literature using Blind Source Separation (BSS) or Independent Component Analysis (ICA) methods rely on a widely used projection techniques family, which is the orthogonal projection. More recently, reported work in other areas of biomedical research is based on the use of oblique projection techniques in order to remove the artifacts from the available recordings and to estimate the contribution of the Signal of Interest (SoI) to the data set. The novel approach presented in this paper is tailored to the problem of foetal ECG (FECG) contribution estimation. The optimal steering vector’s computation procedure for the oblique projection is based on the use of the cyclostationary properties of the SoI and on the concept of Multidimensional Independent Component Analysis (MICA). The latter is used to gather into independent subspaces the components estimated using a BSS/ICA method and corresponding to the foetus’ heartbeats, the mother’s heartbeats and various artifacts and noise sources. Early comparative results with the orthogonal approach obtained after application of both methods to synthetic non-invasive FECG recordings illustrate the reliability as well as the effectiveness of the proposed method. A discussion and perspectives for future research conclude this work. Keywords— foetal electrocardiogram, oblique projection, principal component analysis, multidimensional independent component analysis, cyclic coherence