A Two-Level Iterative Reconstruction Method for Compressed Sensing MRI

Abstract

This paper proposed a novel two-level iterative reconstruction method for compressed sensing magnetic resonance imaging (CS-MRI). For the model of this method, we incorporated the phase correction matrix and region of support (ROS) to guarantee more accurate reconstruction. For the algorithm of the method, we proposed an iterative strategy to reduce memory and computation time, and a two-level strategy to take into account both low and high frequency k-space data separately. Simulation results on normal brain image and angiogram of cerebral demonstrated that the median square error of this proposed method was much less than the traditional method. The error reduction ratios are 11.94% for brain image and 4.53% for angiogram of cerebral, respectively.