Multi-channel MCG signals filtering method based on multivariate variational mode decomposition

With the help of an extremely sensitive magnetic sensors array, magnetocardiography (MCG) can record magnetic field signals produced by the cardiac electrical activity. However, MCG is a very weak magnetic field signal, and tends to be overwhelmed by unsuppressed ambient magnetic noise. To address this issue, this work proposed a multi-channel MCG signals synchronously denoising method based on multivariate variational mode decomposition (MVMD) algorithm. Firstly, the MVMD algorithm extracted spatiotemporal correlation among channels and decomposed multi-channel signals into three-dimensional intrinsic mode functions (IMFs) matrix. By analyzing the time and frequency characteristics of each IMF, we figured out powerline interference (PLI), baseline wander (BW), and white Gauss noise (WGN), and then the sub-band filters removed the noise components before reconstruction. The correlation of three types of noises and MCG signals between channels were totally explored, and both simulated and experimental datasets verified the efficiency. The simulated results indicated that the average reconstruction error of the MCG signals is less than 1 pT. In the human MCG experiment, PLI and BW were reduced by approximately 60 dB and 20 dB, respectively, making the Q, R, and S waves clearly visible. Our work offers comprehensive solutions for denoising multi-channel MCG signals and significantly improves the denoising accuracy.