Electric current source estimation by high-res magnetic field restoration from sparse magnetic measurements

Electric current source estimation (also known as inverse problem) is a common problem to various electric and magnetic imaging applications. For example in magnetocardiography (MCG), electric activities in the heart are reconstructed from sparse measurements of their magnetic field for diagnosis. The inverse problem requires optimizing a highly nonlinear process even in case of a single current source. The existing methods thus often need a good initialization, which cannot be directly provided by the sparse measurements. In this paper we restore the high-res magnetic field from the sparse measurements based on a model learning scheme. By this means a good initialization can be obtained for solving the inverse problem. We then introduce a dual-step optimization algorithm to estimate the current source.