Yilong Liu1,2, Mengye Lyu1,2, Yanqiu Feng1,2, Victor B. Xie1,2, and Ed X. Wu1,2
1Laboratory of Biomedical Imaging and Signal Processing, The University of Hong Kong, Hong Kong, China, People's Republic of, 2Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China, People's Republic of
Synopsis
A navigator-free motion correction method for
Cartesian FSE is proposed for intermittent motion compensation. It first divides
all shots into several motion-free groups, then estimates and corrects inter-group
motion. In vivo imaging results show that this approach has decent
performance in dealing with intermittent motion problems. Though only demonstrated
with Cartesian acquisition, it is also applicable for non-Cartesian acquisitions,
such as spiral and radial acquisitions.Introduction
Patient motion remains to be a common problem that degrades the MR image
quality in clinical diagnosis. In general, patient motion can be continuous (movement
of unsettled pediatric patient), periodic (respiration), or intermittent
(coughing, head shaking) [1,
2]. For intermittent
motion, acquisitions with navigation echoes [3] or inherent navigators [4] have been shown to
be successful in eliminating motion artifacts, but these approaches can prolong
acquisition. Though some navigator-free motion correction methods have been
proposed for Cartesian acquisition, they can only deal with translational
motion [5] or short-lived motion
(twitch, swallowing, etc.) [6]. Here, we propose a motion correction method
for Cartesian Fast Spin Echo (FSE), which can deal with both translation and
rotation. This method first divides the whole data set into several “motion-free”
groups, then estimates and corrects inter-group motion.
Methods
Motion estimation and
correction
(i) Coil Sensitivity Map (CSM) calculation: Images are
first reconstructed without motion correction, and CSMs are generated using these
uncorrected images. Though corrupted with motion, the generated CSMs can still
be used for SENSitivity Encoding (SENSE) reconstruction.
(ii) Motion detection: Motion is detected using a sliding
window approach, as described below. Suppose we use a sliding window with
window width N, i.e., every N successive shots constitute a frame, with which an
image is generated using SENSE. Then inter-frame motion is estimated with image
registration techniques, e.g. those used in PROPELLER [7]. Here, motion is
identified if estimated motion between two frames exceeds certain range. This
approach determines when the motion occurs and correspondingly divides the
whole data set into “motion-free” groups. If some shots are corrupted by
intra-shot motion, or the number of shots in a group is too small to estimate
motion, they are rejected for the final reconstruction.
(iii) Inter-group motion estimation and correction: After
SENSE reconstruction for each group, image registration techniques are used to
estimate inter-group motion. Note that rotation
may cause sampling overlapping or missing in some k-space regions, leading to
data inconsistency. In such cases, non-Cartesian SENSE [8] is performed for image
reconstruction, and low-rank matrix completion is used to further reduce
k-space data inconsistency[9].
Data acquisition
In vivo
T2-weighted data were acquired on a 3T Phillips scanner with an 8-channel head
coil. The shot order was shuffled to make each group distribute more uniformly
in k-space. The shot number was set to 15, with echo train length set to 24,
TR/TE = 3000/115 ms. Two volunteers were asked to stay still during the first
scan (reference scan), and move his/her head during the following 2~3 scans in
an intermittent manner, and the rotation range was within ±10°.
Results
Figure 2 shows images reconstructed from motion-free groups, each
consisting of 5, 6 and 4 shots (from left to right), respectively. Though suffering
from SNR loss and residual aliasing, they can still be used for motion
estimation. Figures 3 and 4 show the final reconstructed images using the
proposed motion correction method for two different volunteers. It can be seen
that higher SNR and less residual aliasing were achieved in these images than
the images reconstructed for each group without introducing motion artifacts.
Discussion and conclusions
In this study, a new method for correcting the
artifacts in MRI caused by intermittent motion is demonstrated. Our results
showed that this method has decent performance when applied to in vivo clinical
imaging settings. For radial acquisitions, researchers proposed to group data
with a center of mass motion detection approach, then estimate and correct inter-group
motion[10, 11]. Though we only demonstrate the effectiveness
of this method for Cartesian acquisition, it can be applied to other acquisition
strategies, such as spiral or radial acquisitions, providing an alternative of
the aforementioned approaches proposed for radial acquisitions. It should be noted
that this approach requires enough shots in each group, thus continuous motion
may compromise its performance.
Acknowledgements
No acknowledgement found.References
[1] M. B. Ooi, et al., MRM 2009; 62(4): 943-954.
[2] C. Malamateniou, et al., AJNR 2013; 34(6): 1124-1136.
[3] R. L. Ehman, et al., Radiology 1989; 173(1): 255-263.
[4] J. G. Pipe, MRM 1999; 42(5): 963-969.
[5] J. Mendes, et al., MRM 2009; 61(3): 739-747.
[6] M. Bydder, et al., MRM 2002; 47(4): 677-686.
[7] J. G. Pipe, et al., MRM 2014; 72(2): 430-437.
[8] R. Bammer, et al., MRM 2007; 57(1): 90-102.
[9] Z. Bi, et al., ISMRM 2013; 2584.
[10] A. Shankaranarayanan, et al., MRM 2001; 45(2): 277-288.
[11] A. G. Anderson, et al., IEEE ISBI 2011; 1528-1531.