Quantitative component selection method using attractor anal-ysis for noise reduction by ICA of magnetocardiogram

Abstract

Magnetocardiogram (MCG) is a noninvasive technique that detects the magnetic field generated by the electrical activity in the heart. Recently, MCG has been attracting a lot of attention in relation to the early detection of heart diseases [1]. As MCG is a multichannel measurement technique and as the signals are not affected by the shape of the lungs and torso, it has high potential for clinical applications [2]. On the other hand, MCG is very weak signal less than 100 pT, and then signal-to-noise ratio (SNR) is very low. Many noise reduction methods are proposed for biomagnetic measurements, particularly independent component analysis (ICA) for separating the signal and noise. However, ICA has the problem that is to distinguish signal or noise components from separated components. In many cases, these is decided by personal qualitative evaluations (knowledge and sense). We have studied a method of environmental magnetic noise reduction using ICA for MCGs. The purpose of this study is to establish quantitative component selection method of the cardiac magnetic field component and the noise component separated by ICA. Therefore, we proposed the method to use attractor analysis and the multiple coefficient of determination.