Gaussian source model based iterative algorithm for EEG source imaging.

Estimation of the neural active sources from the scalp electroencephalogram (EEG) is an ill-posed inverse problem. In this paper, we propose a new source model: Gaussian distributed Source Model (GSM), to model the activations in brain. GSM may imitate an Isolated Source Model (ISM) or a Distributed Source Model (DSM) by adopting different supporting range parameter of the Gaussian function. Using GSM, an iterative Gaussian source Imaging Algorithm (GIA) is developed to detect the EEG sources. As GIA dynamically reduces the solution space, the solution may gradually converge to a desired distribution. A comparative evaluation among LORETA, FOCUSS and GIA was conducted for both isolated point sources and distributed sources, the results demonstrate that GIA is more flexible and efficient for various actual sources configurations. Finally, GSM was applied to real recordings obtained from a visual spatial attention task; the corresponding source activation areas of the early component are localized in contralateral occipital cortices, consistent with the retinotopic organization of early visual spatial attention effects.