Rethinking MRI random signals modeling

Abstract

Based on both the Physics of MRI and the central limit theorem, it is common practice to assume that the noise in MR images is Gauss distributed, but from an MR signal post-acquisition standpoint, this modeling approach can be proved to be erroneous, especially when the SNR is low. In this article, we present a thorough analysis that shows why the Gaussian model was adopted, and through the MR complex raw data post-acquisition mathematical treatment, the Rician model will be developed and proved to be the right MR random signals model.