Highly accelerated parallel imaging methods for localized massive array coils: comparison using 64-channel phased-array data

Massive parallel phased-array systems with 32, 64, or more receive channels have potential to achieve high scan time saving for parallel MRI. In the past decade, a number of parallel MRI methods have been proposed including SENSE, SMASH, PILS, GRAPPA, SPACE-RIP, SEA, and other methods. In this work, we investigate the optimality of four reconstruction methods for parallel imaging with massive localized linear phased-array coils. In particular, the artifact power, SNR, and scan efficiency of different methods are systematically compared and analyzed. The studies are based on real MR data collected using a 64-channel system for 2-D MR imaging. The results show that the auto-PILS method and the improved GRAPPA method provide the best SNR and minimal artifact. In addition, the auto-PILS method is most efficient in scan time reduction. Interestingly, SMASH is not optimal for use with the highly localized coils with coil width on the same order as the voxel size