Abstract #3652

Sngas J, Eggers H, Knopp T

A new, non-iterative algorithm is proposed for the reconstruction of non-Cartesian acquisitions in parallel imaging. It performs the reconstruction in k-space, where each sample is estimated by a linear combination of only a subset of the acquired data. The weights of the linear combination are derived from the coil sensitivities using Fourier transforms and linear algebra. It is shown that the theoretical mean square error of the reconstruction is a useful guideline to locally select the subset of the acquired data for the estimation. The algorithm is demonstrated on simulated and measured radial and spiral acquisitions.