A fast approach to processing coil sensitivity maps using a neural network
Boesiger P, Buehrer M, Kozerke S
Obtaining information about coil sensitivities is crucial when using sensitivity encoding. Raw sensitivity maps, which can be calculated by dividing single-coil images by the body-coil image, are often impaired by noise, especially in regions with low spin density (e.g. in the lungs). In this work a neural network is used for fast fitting and extrapolation of the raw sensitivity data. Using this approach the computation time for processing the raw sensitivity maps could be reduced by a factor of twelve relative to polynomial fitting approaches.