Regularizing GRAPPA using simultaneous sparsity to recover de-noised images

Abstract

To enable further acceleration of magnetic resonance (MR) imaging, compressed sensing (CS) is combined with GRAPPA, a parallel imaging method, to reconstruct images from highly undersampled data with significantly improved RMSE compared to reconstructions using GRAPPA alone. This novel combination of GRAPPA and CS regularizes the GRAPPA kernel computation step using a simultaneous sparsity penalty function of the coil images. This approach can be implemented by formulating the problem as the joint optimization of the least squares fit of the kernel to the ACS lines and the sparsity of the images generated using GRAPPA with the kernel.