Dipole Source Localization in Magnetocardiography

Dipole source localization (DSL) is complicated as it involves solving electromagnetic inverse problems in Magnetocardiography (MCG). The accuracy of the inverse solutions depends upon the adequacy of the forward model and the DSL algorithm used. We have developed two simplified models based on double magnetic dipoles source and a single current dipole source respectively. In this paper, the inverse solutions of DSL using Levenberg-Marquart (LM) algorithm and Nelder-Mead algorithm (NM) were discussed. Numeral simulations were performed using the MCG data generated from two different graphical models, which were reconstructed in terms of the parameter character of the real MCG measurements. An analysis of the average localization error in the ST-T segment at different noise levels and with different initial values was given. Our results indicated that the NM algorithm for both models works effectively in the case of initial values having 10% random noise. Furthermore, the localization accuracy less than 1 mm has been achieved in our research.