A New Combination Scheme of GRAPPA and Compressed Sensing for Accelerated Magnetic Resonance Imaging

Abstract

Both compressed sensing (CS) and parallel imaging (PI) can be used to accelerate magnetic resonance imaging (MRI) by under-sampling the k space data. Several methods combining CS and PI have been proposed to further improve the scanning speed. In this paper, we proposed a new approach to combine CS and PI. We used GRAPPA (Generalized Autocalibrating Partially Parallel Acquisitions) algorithm to reconstruct local under-sampled k space data, and CS to reconstruct the whole k space data for each coil. In the CS reconstruction step, we constrained that the reconstructed k space data should be assimilated to both the sampled k space data and the reconstructed k space data by GRAPPA. In addition, we designed a new sampling strategy to improve the quality of image reconstruction. In vivo imaging results demonstrated that the proposed approach could effectively remove artifacts and improve the image quality.

Keywords

• magnetic resonance imaging (MRI)
• compressed sensing (CS)
• parallel imaging (PI)
• sparse sampling