3D TV-Based Compressed MR Image Reconstruction Using a Primal Dual Algorithm

Abstract

Emerging trends toward fast magnetic resonance (MR) imaging focus on partial Fourier measurements and k-space undersampling. However, undersampling inevitably violates the Nyquist sampling criterion, thereby Fourier reconstructions exhibit aliasing artifacts and reduced signal-to-noise ratio (SNR). By establishing a direct link between sampling and sparsity, compressed sensing (CS) has, however, made it possible to accurately reconstruct MR images from undersampled k-spaces [1]. In this study, we introduced a fast primal-dual (PD) algorithm for CS MR image reconstruction using variable-density spiral trajectories. The algorithm was developed for 3D total variation (TV) and Huber regularizations and was found to be very efficient and promising for fast 3D MRI.