Distributed Source Reconstruction Using a Non-Linear Spatio-Temporal Regularization Method: An Alternative to LORETA

Abstract

In the EEG / MEG inverse problem solving, it appears that most of the time single dipole models fail to produce acceptable results — specially with the late components of evoked potentials — when space and time overlapping in the source pattern is very likely. At the same time, minimum-norm multi-source solutions exhibit over-smoothed magnitude patterns and, unless specified, do not respect brain anatomical constraints like sulcus borders. Moreover, these methods fail to recover deeper sources because dipoles located at the surface of the source space (the cortex surface, basically) would be privileged for the same data set. Recently Pascual-Marqui et al. have presented a generalized minimum-norm Solution to the inverse problem that appears to overcome some of these limitations but still offers smoothed dipole images [1].