Solution for MEG inverse problem using Signal Space Separation and Magnetic Field Tomography

Magnetic Field Tomography (MFT) is a source localization method for Magnetoencephalography (MEG), a non-invasive method to observe the brain activity. MFT just requires the source to be a linear combination of lead fields that describe the distribution of the sensitivity of each sensor, while other commonly used MEG source localization methods such as equivalent current dipole (ECD) fitting or the beamformer require some more inappropriate assumptions. However, less requirements on the source results in a huge amount of computational load in MFT. In this paper, the reduction of the computational load for MFT was achieved by considering the coefficients of multipolar expansion as the measurements of virtual sensors. These coefficients are obtained by performing Signal Space Separation (SSS) in which the exclusion of external magnetic field generated by the external sensor arrays is enabled. Based on our simulation, the calculation time was reduced from 6 hours to about 10 seconds preserving the source localization ability.