用于心脏电活动成像的空间滤波器输出噪声抑制方法

利用人体表面测量的心脏磁场数据无创成像心脏电活动,需要解决的关键问题是提高其重建分布电流源偶极矩强度的分辨率.本文在最小方差波束成形(MVB)方法的基础上,提出了一种可抑制空间滤波器输出噪声功率增益(SONG)的波束成形方法,目的是通过构造一种新的滤波器权矩阵,约束空间滤波器的输出噪声功率增益,提高重建分布电流源偶极矩强度的分辨率,即分布电流源空间谱估计的源分辨能力,从而增强心脏电活动磁成像的分辨率.文中给出了电流源重建的理论分析和仿真结果;比较了该方法与MVB方法的差别;并给出了两个健康人36通道心脏磁场数据的电活动成像.结果表明,SONG方法分辨电流源的能力较强,能够观察到心脏电磁场信号R峰时刻健康人的心室电活动较强,心房电活动较弱等特征.

The method of suppressing spatial filter output noise-power gain for cardiac electrical activity imaging

For non-invasive imaging of cardiac electrical activity using magnetocardiogram (MCG) data measured on human body surface, a key problem that needs to be solved is to enhance the spatial resolution of reconstructing distributed current source dipole moment strength in MCG imaging. In this paper, a beamforming method of suppressing spatial filter output noise-power gain (SONG) is proposed based on the minimum variance beamforming (MVB). The purpose is to improve the resolution of the distributed source dipole moment strength reconstruction, i.e., the ability to resolve the source for distributed current source spatial spectrum estimation, in order to enhance the resolution of the cardiac electrical activity magnetic imaging. The method offers a new spatial filter weight matrix by using a low-trace positively-semidefinite matrix that will affect the spatial filter output power, on the premise that the influence of noise spatial spectrum of spatial filter on the estimation of current source spatial spectrum has been constrained by the noise spatial spectrum intensity normalization. The positively-semidefinite matrix is specially constructed to satisfy the condition that the eigenvalue is not greater than 1 and the trace of the matrix is lower than its order, so that it can be used to constrain the spatial filter output noise-power gain for improving the robustness to noise of the source spatial spectrum estimation. In addition, a classical model of the horizontally layered conductor is used as the heart-torso model to calculate the lead-field matrix that needs to be used in source spatial spectrum estimation. The results obtained in this study are as follows. For validating the proposed method, a theoretical analysis and simulation tests of the current source reconstruction are performed, where the SONG and MVB methods are compared and a parameter of the signal-to-noise ratio is considered according to the realistic MCG data. In this paper we also give the cardiac electrical activity imaging of 36-channel cardiac magnetic field data of single-cycle from two healthy people, where a heart profile from the magnetic resonance imaging is used as a reference and adjusted to the MCG measurement system. The results show that the SONG method has ability to better resolve the current source and can observe the significant electrophysiological characteristics such as the strong electrical activity in the ventricles of the healthy people at the time of R_peak. In summary, our proposed method can improve the visual effect of the cardiac electrical activity imaging, when the signal-to-noise ratio of the single-cycle cardiac magnetic signal is not lower than 10 dB. Therefore, this method of measuring the non-invasively imaging cardiac electrical activity is a promising one and helpful for relevant medical research and applications.