Noise behavior in gridding reconstruction

Abstract

The paper addresses the properties of the noise in gridding reconstruction, an algorithm for reconstruction from nonuniform samples. Sequences with time-varying gradients, such as spiral or projection reconstruction (PR) techniques, are being increasingly used in magnetic resonance imaging (MRI). Since these techniques sample k-space nonuniformly, some kind of algorithm is needed to map the data onto a Cartesian frame to allow an inverse Fourier transform through an FFT. The authors present an analytical characterization of the image noise after gridding and inverse Fourier transform for the most popular sampling techniques used in MRI.