Magnetocardiography Signal Reconstruction With Reduced Source Space Based on Current Source Variance

In this study we use magnetocardiogram (MCG) signals to perform the current source reconstruction of the signal generators. The source model used is a distributed source model that allows visualizing the current flow in the human heart. The heart volume is divided into voxels, each one being the site of one current dipole that has varying orientation and magnitude. We then perform a variance analysis of the dipole components waveforms to detect which dipoles carry the most significant information. Since every dipole has a fixed position, the localization of the ones that mainly contribute to the field generation allows identifying in what portion of the source space region the activation wavefront is mainly localized. After excluding the dipoles with the less information content, the forward field computation is performed to check whether the magnetic signals differ significantly from the measured data. The method is applied on the MCG of a patient affected by Wolff-Parkinson-White syndrome to localize the accessory conduction pathway in terms of current density. The comparison of the results obtained between vector and normal field components data allows improving the results when using vector data.