MCG source reconstruction based on greedy sparse method

Abstract

Current source reconstruction, i.e., reconstructing current dipole distribution through measured array signals of cardiac magnetic field on body surface, is a method for non-invasively study on the heart electrical activity. In this paper, the relationship between measured magnetic signals and current dipole distribution is described by a linear equation, and a sparse solution of current source reconstruction is achieved using a fast greedy method. This method can significantly decrease the computational complexity of or- thogonal matching pursuit (OMP) algorithm by means of approximating orthogonalisation and improving the selection vector strategy per iteration. Thereby, the sources with large dipole strength can be fast searched out with high accuracy. A set of magnetocardiogram (MCG) data of normal subject is used to demonstrate the effectiveness of this method that the trajectory of reconstructed dominant sources, whose strengths are more than 65%, is almost consistent with conduction process in depolarization and repolarization. The average goodness of fit (GOFs) of measured MCG and the magnetic field map generated by the reconstructed current sources during QRS complex and ST-T segment are 99.36% and 99.78%, respectively.