磁共振成像PROPELLER采样数据重建中的运动估计新算法

A New Algorithm for Motion Estimation in the MRI Reconstruction of PROPELLER Sampled Data

PROPELLER(推进器)采样技术能够利用K空间中心重叠采样区域的数据来估计采集过程中受检查者的运动进而加以补偿,对运动伪影的消除效果非常显著。然而,由于其重建时的运动估计是基于最大化频域空间上相关系数的配准算法,该算法为了实现旋转估计与平移估计的分离,在进行旋转估计时,仅仅采用K空间数据的模,在数据量有限的情况下造成估计精度较低,在重建图像上表现为模糊及星条状伪影。本研究基于最大化图像空间上的互信息提出一种PROPELLER采样数据的运动估计新算法,首先由每个K空间带进行傅立叶逆变换后取模重建出系列临时图像,对这些图像进行模糊增强后以互信息作为相似性测度迭代搜索最优的运动参数。实验证明,该方法能显著提高PROPELLER采样数据重建中运动估计与补偿的精度,从而更好地消除伪影,特别是用于有运动时T1加权头部成像时。

PROPELLER （Periodically Rotated Overlapping ParallEL Lines with Enhanced Reconstruction） MRI, can extract inter-strip motion information from data in central overlapped sampling area, it offers an effective means to suppress the artifacts caused by patients＇ rigid motion. However, the current algorithm for motion estimation proposed by Pipe is actually the registration by maximizing correlation energy in k-space space （frequency domain） and only magnitudes of complex k-space data are taken into account while performing rotation estimation to exclude the influence of translation. Additionally, little data can be obtained in the central overlapped area of PROPELLER sampling because each strip must be collected quickly as possible. Therefore, the algorithm produces image with low quality and being contaminated by steak artifacts due to the imprecise estimation of motion parameters as shown in this paper. In this work, the algorithm based on the registration by maximizing mutual information in image space （spatial domain） was proposed for estimating motion parameters. The algorithm overcome the limitations of current algorithm and produced an improved accuracy. The algorithm exhibited the performance in reducing motion artifacts and improving quality of image reconstructed from PROPELLER sampled data, especially when implemented on T1-weighted MRI data of head.