Current Dipole Estimation in MEG by Spatial Interpolation of Magnetic Sensors

Abstract

In this study, a method for current dipole estimation with spatial interpolation of measured magnetic fields in magnetoencephalogram (MEG) is proposed. This method applies virtual magnetic (M _v ) sensors uniformly interpolated in the surface comprised of MEG (M _r ) sensors. Orientations and output signals of the virtual magnetic sensors are calculated by a spatial linear interpolation based on those of real magnetic sensors. For determining an optimal value of regularization parameter (α) in minimum norm inverse solutions, M _v sensors are divided into some groups randomly. Quantities of magnetic field (b _k ) at the M _r sensors are calculated by estimated dipoles with each group of M _v sensors. α is determined by minimization of the total sum of differences between b _k and the output signals measured by M _r sensors. In order to confirm the validity of the proposed method, we estimated source distribution of N20m components of unilateral and bilateral somatosensory evoked fields (SEFs) measured by a 160-channel axial SQUID gradiometer system, when subject’s right and/or left median nerve was electrically stimulated. The proposed method was able to estimate source distribution in the primary somatosensory cortex (S1). Furthermore, the source distribution estimated by the proposed method tended to be more confined than that estimated by a cross validation error method. These results indicate that the virtually improved spatial resolution of MEG data can estimate more confined source distribution.