Application of neural network approach for MAP-MRF model-based MEG source imaging

Magnetoencephalography (MEG) source image reconstruction is an important and difficult problem in image processing applications. It can be formulated as an inherent ill-posed and highly underdetermined linear inverse problem, encompassing a great variety of signal modeling and processing methods. According to anatomical and physiological knowledge, source distributed image is modeled by a Markov random field (MRF), and the reconstruction is defined as the maximum a posteriori estimate (MAP) based on a Bayesian framework. To acquire a global minimum solution from the posterior energy function, many optimization methods (such as GA, SA, MFA, etc.) appear in the literature are investigated. In this paper, we incorporate the ideas of artificial neural networks into this difficult optimization task. Several computer experiments were conducted in order to assess the performance of the introduced technique.