基于多通道图像相关性的改进GRAPPA算法

An improved GRAPPA algorithm based on the correlation between multi-coil images

针对传统GRAPPA(generalized auto-calibrating partially parallel acquisitions)算法存在随欠采样倍数增加重建图像质量下降的问题,提出一种改进算法.该算法利用多通道图像间的更多相关信息对传统GRAPPA算法进行改进,突破传统GRAPPA算法仅利用自校准信号进行权重系数估计的局限,并进一步使用已采集数据之间的一般相关性.实验结果表明,改进后的GRAPPA算法能重建出更高质量的磁共振图像.

Conventional GRAPPA （generalized auto-calibrating partially parallel acquisitions） algorithm uses the auto-calibration data of additional acquisition to fit the missing K-space data and reconstruct desired image from a multi-coil under-sampling data set. However, as the acceleration factor increases, reconstruction quality decreases quickly. To address this issue, we propose a new reconstruction algorithm to improve the conventional GRAPPA by taking more correlation information of multi-coil images. It overcomes the limit of conventional GRAPPA which only uses the auto-calibration data to estimate the fitting coefficients. It takes the available relationship of all the data making reconstructed the better-quality image. Experimental results show that the proposed method could provide a better reconstruction than conventional GRAPPA.