After performing noise reduction via preprocessing, sensor-level data, especially those from planar gradiometers, may indicate the probable number and approximate locations of active sources. In order to actually locate the sources, several different unique solutions to the ill-posed electromagnetic inverse problem exist. Each localization technique has its own modeling assumptions and thus also strengths and limitations. Therefore, the MNE software provides a selection of inverse modeling approaches. Importantly, in all of the approaches discussed here, the elementary source employed is a current dipole, motivated by the physiological nature of the cerebral currents measurable in M/EEG (Hämäläinen et al., 1993). Different source modeling approaches are set apart by the selection constraints on the sources and other criteria to arrive at the best estimate for the cerebral current distribution as a function of time.
