Inverse reconstruction of left ventricular action potentials from contactless magnetocardiography enables non invasive high-resolution mapping of repolarization heterogeneity

Abstract Contactless magnetocardiographic mapping (MCG), feasible also in unshielded (un-) hospital environments, offers novel alternatives for noninvasive clinical cardiac electrophysiological assessment [1]. A mathematical method to reconstruct the ventricular action potential waveform (VAPw) from MCG-derived pseudo-current arrow mapping (CAM) [2] was recently applied to clinical un-MCG recorded during interventional electrophysiology (IEP) and validated by comparison with simultaneous right ventricular monophasic action potential (RVMAP) recordings [3]. However, that study revealed limitations in the spatial assessment of local ventricular repolarization (VR) dynamics. Objective To develop a different approach based on the inverse 3D calculation of the primary current density (PCD) of left ventricular (LV) tissue from clinical un-MCG, and to assess its performance in improving the spatial accuracy of VAPw reconstruction. Methods un-MCG was acquired in the IEP laboratory with a 36-channel DC-SQUID system (sensitivity 50 fT/Hz in the DC–250 Hz bandwidth). The MCG S/N ratio was optimized by signal averaging. A 3D multidomain mesh was generated, comprising an empty ellipsoidal left ventricle, an intraventricular blood pool, and the surrounding torso (Figure 1). An anisotropic Minimum Norm Estimate (MNE) solver [4], [5] was applied to reconstruct the PCD, explicitly accounting for fiber direction–dependent myocardial conductivity [6], [7]. Subsequently, a localized adaptation of the pseudo-CAM method [2] was used to derive the action potential waveform (LVAPw) directly from the PCD along each fiber and to compare it with simultaneously recorded RVMAP (Figure 2). Results LVAPw reconstruction from the MNE-PCD of un-MCG provides discrete, node-specific localization at ~2 mm tetrahedral resolution. The transition from averaged cardiac current estimation using pseudo-CAM [2] to localized PCD analysis improves spatial resolution, enabling independent evaluation of depolarization and repolarization across small myocardial bundles. The evident waveform and duration differences between LVAPw and RVMAP (Figure 2A) underscore sensitivity in detecting local VR inhomogeneity. A few, yet still evident, instances of incomplete VR were successfully corrected by a specifically designed automated algorithm (Figure 2B), unless occurring at nodes with excessively high noise (Figure 2C). Conclusions The results obtained with MNE-localized reconstruction of LVAPw suggest that the incomplete repolarization previously observed in CAM-derived waveforms is likely a localized artifact due to the interplay among algorithmic spatial smearing, null-space projections, and environmental noise, severely attenuating the calculated PCD at specific nodal sites. Targeted advancements in 3D vector sensor technology and nonlinear mathematical modeling will directly address current physical constraints, significantly enhancing the diagnostic accuracy of the MCG method.Figure 1.3D multidomain mesh modelFigure 2.Comparison betwee LVAPw and RVMAP