基于自一致性的磁共振并行成像高效重构算法

磁共振并行成像技术能够显著地减少成像时间,然而高质量的图像重构比较困难.为了提高重构图像的质量,基于自一致性的SPIRiT框架,提出了一种高效的重构算法.该算法针对一个含有数据一致性、校准一致性和联合稀疏性正则项的复杂优化问题,首先将该问题简化成一般性最优化问题;再使用算子分离算法将其分解成一个梯度计算问题和一个可通过软阈值法求解的去噪问题;最后,再使用加速方案对算法进行加速.实验结果表明,当加速因子为8时,采用所提出的新算法的重构图像比采用POCS算法的重构图像的SNR提高约2.4,dB,且重构时间也节约了约30%.对于要求高质量重构图像的场合,所提出的算法能够满足需求.

Efficient Reconstruction Algorithm for Parallel Magnetic Resonance Imaging Based on Self-Consistency

Parallel magnetic resonance imaging （PMRI） can significantly reduce the imaging time, but it is difficult to reconstruct a high quality image. Based on the iterative self-consistent parallel imaging reconstruction （SPIRIT） framework, an efficient algorithm was proposed for improving the quality of reconstruted image. This algorithm firstly simplified the problems with data consistency, calibration consistency and a regularization term into a general optimization problem. By using the operator splitting technique, the optimization problem was decomposed into a gradient problem and a denoising problem which could be solved easily by using the soft-thresholding method. Then, the proposed algorithm was accelerated. The experimental results demonstrate that when the accelerating factor is 8, the proposed algorithm can not only achieve significant signal-to-noise ratio （SNR） improvement （up to 2.4 dB） compared to the projection over convex sets （POCS） algorithm, but also save 30% of the reconstructed time. The proposed algorithm is particularly suitable for the high quality reconstruction of PMRI.