Rejection of Wire-induced Magnetic Noise Using Wavelet Transformation and Independent Component Analysis for Magnetocardiogram

摘要

We present a magnetic noise rejection method using wavelet transformation (WT) and independent component analysis (ICA) for analyzing magnetocardiograms (MCGs) . In the ICA, when the signal-to-noise (SN) ratio of the data is low, the accuracy of the noise rejection becomes low and the determination of the contraction dimension becomes difficult. For example, MCG has a very low SN ratio in patients who have a wire surgically fixed in, their sternum. To solve this limitation, we have proposed a method to distinguish signal and noise components using Mahalanobis distance and WT as pre-processing for ICA. We carried out a simulation which attached a wire to a normal subject in order to show accuracy improvement of the noise rejection by the proposed method. The SN ratio of measured MCG was the 6dB. The correlation coefficient of waveform after noise rejection with control data (waveform without the wire noise) was calculated. As this result, it became 0. 98 high value. The proposed method was effective for the wire noise rejection.