A comparative study of compressed sensing approaches with splitting Bregman framework for radial UTE MRI

This paper investigates two compressed sensing (CS) approaches that can be used to reconstruct radial Magnetic Resonance (MR) images with undersampled k-space measurements. Combining CS with gridding and non-uniform fast Fourier transform (NUFFT), yields two different approaches: Regridding-CS and NUFFT-CS. Under splitting Bregman framework, these approaches can decrease the load of data acquisition while recovering MR images through l₁-norm and total variation (TV) optimization. Experiments using a phantom example have verified that the NUFFT-CS achieves better image quality than the Regridding-CS.