Biomagnetic localization of electrical current sources in the human heart with realistic volume conductors using the...

The boundary element method was applied in order to investigate the localization accuracy for focal sources measured from MCG data. Various homogeneous volume conductor models were composed: the individually shaped torso, a scaled standard torso, an unscaled standard torso, a scaled cuboid and a scaled ellipsoid. We implemented these models in single-dipole inverse solution techniques. High resolution multichannel data were analysed from two patients showing ventricular extrasystoles and two patients suffering from Wolff-Parkinson-White syndrome. Moreover, we report the localization of shallow- and deep-lying catheters (depth 9 cm and depth 17.5 cm below the measurement grid). Using an individually shaped homogeneous torso yields a localization error of less than 3 cm even for the deepest sources (mean error 2.4 cm). Probability-based dipole localization shows that the remaining error could only partly be explained by data noise statistics. Therefore it seems to be due to either inner inhomogeneities or the inadequacy of the single current dipole or a combination of the two. Thus clinically useful localization accuracy in the millimetre range requires more sophisticated volume conductor and source models. The evaluation of measurement data and simulation study shows that a scaled cuboid model can provide nearly the same localization accuracy as the individually shaped torso model. Single dipole reconstruction with this model is computationally faster than that with the individually shaped model of the human body and is fast enough for use in clinical applications.