Combination of Canonical Correlation Analysis and Empirical Mode Decomposition Applied to Denoising the Labor Electrohysterogram

Abstract

The electrohysterogram (EHG) is often corrupted by electronic and electromagnetic noise as well as movement artifacts, skeletal electromyogram, and ECGs from both mother and fetus. The interfering signals are sporadic and/or have spectra overlapping the spectra of the signals of interest rendering classical filtering ineffective. In the absence of efficient methods for denoising the monopolar EHG signal, bipolar methods are usually used. In this paper, we propose a novel combination of blind source separation using canonical correlation analysis (BSS_CCA) and empirical mode decomposition (EMD) methods to denoise monopolar EHG. We first extract the uterine bursts by using BSS_CCA then the biggest part of any residual noise is removed from the bursts by EMD. Our algorithm, called CCA_EMD, was compared with wavelet filtering and independent component analysis. We also compared CCA_EMD with the corresponding bipolar signals to demonstrate that the new method gives signals that have not been degraded by the new method. The proposed method successfully removed artifacts from the signal without altering the underlying uterine activity as observed by bipolar methods. The CCA_EMD algorithm performed considerably better than the comparison methods.

The Use of Ensemble Empirical Mode Decomposition With Canonical Correlation Analysis as a Novel Artifact Removal Technique

“…[26]. This method assumes that the autocorrelation at lag one of the artifact components differs from that of the true desired signal.…”

Effect of filtering on the classification rate of nonlinear analysis methods applied to uterine EMG signals

“…It also includes the use of the piecewise stationary pre-segmentation approach [21] to increase the accuracy of the synchronization measures and denoising the monopolar EHG signals using the combined Canonical Correlation Analysis and Empirical Mode Decomposition [22] to enhance the spatial resolution. The main outcome of this study is to urge caution and systematic exploration of filtering when analyzing EHG nonlinearity to increase pregnancy/labor classification rate using uterine EMG signals.…”

Graph analysis of uterine networks using EHG source connectivity

“…The data were recorded at the Landspitali university hospital (Iceland) using a protocol agreed by the relevant ethical committee (VSN02-0006-V2) and at the Center for Obstetrics and Gynecology (Amiens, France), using a protocol agreed by the relevant ethical committee (ID-RCB 2011-A00500-41). The EHG signals were manually segmented and denoised by using a CCA-EMD method developed in our team [13]. After segmentation and denoising, we obtained 183 labor and 247 pregnancy EHG bursts, associated to contractions, at different weeks of gestation.…”