Magnetoencephalography (MEG) is a multi-channel imaging technique. It measures the magnetic field
produced by electric currents inside the brain via an array composed of a large number of superconducting
quantum interference devices sensors. These measurements are then used to estimate the location and the
strength of those electric currents. The estimated quantities are superimposed with anatomic magnetic
resonance images via coregistration to create the magnetic source images. This paper attempts to present
a theoretic framework of MEG beamforming for magnetic source imaging by aiming at some fundamental
issues and providing several new findings in the five related aspects: (i) physical concept, (ii) mathematical
formulation, (iii) statistical description, (iv) beamforming principle, and (v) magnetic source images.