Abstract:

This paper reviews the commonly-used image reconstruction algorithms in parallel magnetic resonance imaging. First, the reconstruction algorithms used in parallel MRI without acceleration was discussed. Then, the commonly-used reconstruction algorithms in parallel MRI with acceleration, SENSE, PILS, SMASH and GRAPPA, were compared. It was shown that, with the same accelerating factor, the quality of GRAPPA and SENSE reconstructions are the best among the four, while that of PILS reconstruction is the worst. In terms of reconstruction speed, SMASH is the fastest, and GRAPPA is the slowest. The performance of all four reconstruction algorithms degraded with increasing accelerating factor, suggesting that increase of imaging speed is at the expense of the cost of SNR. We also implemented the reconstruction algorithms for parallel imaging on an Matlab GUI platform. The resulting software has the functions of reading raw data, showing sampling trajectory, calculating coil sensitivity, choosing image reconstruction method, analyzing and comparing image quality etc. We believe the software will be helpful for study and research on parallel MRI reconstruction algorithms.

Key words: MRI image reconstruction, k-space data, parallel imaging