Zhe Zhang1, Xiaodong Ma2, Lanxin Ji1, Jing Jing1, Wanlin Zhu1, Zhangxuan Hu3, Yishi Wang4, Hua Guo3, and Yongjun Wang1,5
1China National Clinical Research Center for Neurological Diseases, Beijing Tiantan Hospital, Capital Medical University, Beijng, China, 2Center for Magnetic Resonance Research, University of Minnesota, Minneapolis, MN, United States, 3Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, China, 4Philips Healthcare, Beijing, China, 5Department of Neurology, Beijing Tiantan Hospital, Capital Medical University, Beijing, China
Synopsis
Multi-shot acquisition
enables high-resolution diffusion imaging, but the artifacts caused by
shot-to-shot phase variation must be corrected. Self-calibrated multi-shot DWI methods utilize parallel imaging reconstruction to solve the phase of each
shot. Previously reported self-calibrated GRAPPA with a compact kernel
(SC-ckGRAPPA) method is compromised by the high reduction factor when recovering
the navigator information. In this work, PM-SC-ckGRAPPA was introduced with the
phase-matched reconstruction, and evaluated via in-vivo experiment. Results
show that PM-SC-ckGRAPPA provides improved reconstruction compared with
conventional approaches, and PM-SC-ckGRAPPA can be a potential tool for
high-resolution diffusion imaging.
Purpose
Multi-shot EPI (MS-EPI) enables higher
resolution diffusion weighted imaging (DWI) than single-shot EPI (SS-EPI), but
the ghost artifacts caused by shot-to-shot phase variation must be corrected.
Some k-space correction methods like ckGRAPPA (GRAPPA with a compact kernel)
for interleaved MS-EPI DWI have been proposed, utilizing the phase variation
from navigator acquisition as additional encoding power[1], to recover the missing
data in k-space of each shot. Self-calibrated method SC-ckGRAPPA has also
been proposed to avoid the mismatch between image and navigator echo and saves the scan
time of navigator acquisition[2]. However, SC-ckGRAPPA is still compromised by
the high reduction factor (R, equals the number of shots) when recovering the
navigator information of each shot. Recently phase-matched reconstruction has
been proposed to improve the performance of SS-EPI DWI GRAPPA reconstruction[3].
In this work, a phase-matched SC-ckGRAPPA reconstruction algorithm (PM-SC-ckGRAPPA)
for MS-EPI DWI is proposed and evaluated using in-vivo experiment. This work
aims to improve the self-calibrated k-space-based MS-EPI DWI reconstruction.Theory
Conventional
GRAPPA: For
conventional GRAPPA DWI reconstruction, each shot is reconstructed by GRAPPA
separately using the convolution kernel
calibrated from b = 0 multi-coil full k-space (assuming no phase variation in b
= 0 acquisition). For an acquisition from m coils, GRAPPA calibration/interpolation
performs on m channels. Then the GRAPPA reconstructed images from each
shot are averaged to generate DWI[4].
Conventional SC-ckGRAPPA: As described in previous work on ckGRAPPA[2],
the multi-coil and multi-shot k-space data are integrated for k-space
interpolation. Figure 1C illustrates GRAPPA and ckGRAPPA k-space interpolation
process. Different from ckGRAPPA which relies on the navigator echo acquisition
for calibration, the SC-ckGRAPPA calibrates on the GRAPPA recovered
full k-space of all shots. For an n-shot with m-coil
acquisition, SC-ckGRAPPA works on n×m channels.
Proposed PM-SC-ckGRAPPA: Previous works show the update of smoothed motion-induced
phase variation improves DWI reconstruction, especially on MS-EPI with
SENSE-like reconstruction[5,6] and SS-EPI with GRAPPA reconstruction[3]. In
this work, the phase-matched reconstruction is incorporated in SC-ckGRAPPA. Different
from SC-ckGRAPPA which directly uses GRAPPA results for calibration[2], PM-SC-ckGRAPPA
calibrates on phase-matched synthetic data from b=0 multi-coil images and DWI smoothed
phase variation, which is performed below: 1) the DWI k-space data for each shot recovered
by GRAPPA are inverse Fourier transformed and SENSE-combined to generate
navigator images; 2) the navigator image for each shot is smoothed using 2D Hann
window filter, and the smoothed complex phase is extracted using cpxPha
= nav / abs(nav); 3) the complex phase for each shot and b=0
image from each coil are combined using complex multiplication to generate phase-matched
image data; 4) the phase-matched image data are Fourier transformed to generate
the multi-shot, multi-coil phase-matched k-space data for PM-SC-ckGRAPPA reconstruction.
Figure 1B illustrates the reconstruction of the phase-matched calibration data.
The proposed PM-SC-ckGRAPPA reconstruction pipeline is shown in Figure 1A.Methods
In order to compare different reconstruction approaches, 6-shot interleaved-EPI
DWI data were acquired on a healthy volunteer on a 3T scanner (Philips, Best, The
Netherlands) with an 8-channel head coil. FOV=210×210mm2, voxel size = 1×1×4 mm3, 20 slices with gap =
1mm, TE = 86 ms (set to minimum), TR = 3500 ms, no partial Fourier, 6 diffusion
encoding directions using b=1000 s/mm2 were acquired with repetition
= 2. On this dataset, three different reconstruction methods were implemented
and compared: a) GRAPPA,
b) SC-ckGRAPPA, c) proposed PM-SC-ckGRAPPA. For all ckGRAPPA k-space calibration and interpolation,
the kernel size was 3×5 (kx×ky) and λ=1e-6 for calibration regularization. The noise map was calculated
from the standard deviation of the difference between 2 repetitions of all
encoding directions, namely NoiseMap = StdDiffusionEncoding(DiffRepietition(DWI)).
The image SNR was calculated as [7] in the brain region for each encoding
direction and each slice.Results and Discussion
Figure
2 shows the two typical slices of the 6-shot data reconstruction. SC-ckGRAPPA shows lower noise level than GRAPPA,
and PM-SC-ckGRAPPA shows minimum noise and signal dropout levels compared with
the other two methods. PM-SC-ckGRAPPA recovers the signal dropout compared with
the conventional methods. The mean image from all diffusion encoding directions
shows the potential of PM-SC-ckGRAPPA to be a new approach for
high-resolution DWI. Figure 3 shows the phase images of 3 typical shots,
reconstructed by GRAPPA, extracted in the phase-match process and reconstructed
by PM-SC-ckGRAPPA, respectively. Compared with GRAPPA results, the phase of the PM-SC-ckGRAPPA
shows reduced noise and artifacts. The mean image SNR of GRAPPA, SC-ckGRAPPA
and PM-SC-ckGRAPPA is 6.67, 7.03 and 7.77 respectively. PM-SC-ckGRAPPA provides
significantly higher SNR over other conventional methods (p < 0.001). Table 1 shows
the averaged SNR of all encoding directions from all 20 slices.
In this work, PM-SC-ckGRAPPA
was introduced and evaluated for interleaved-EPI DWI with human brain
acquisition. The strategy can be extended to multi-band and 3D multi-slab
acquisitions, and the application scenario can be extended to other organs. As
reported in previous works, the GRAPPA-based approach might be more robust than image-based reconstruction in DWI
when subject motion occurs[4]. Conclusion
The proposed PM-SC-ckGRAPPA
provides improved reconstruction compared with conventional GRAPPA and SC-ckGRAPPA.
The interleaved MS-EPI acquisition with PM-SC-ckGRAPPA reconstruction can be a
potential approach for high-resolution DWI.Acknowledgements
No acknowledgement found.References
[1] Ma X, Zhang Z, Dai E,
Guo H. Improved multi-shot diffusion imaging using GRAPPA with a compact
kernel. Neuroimage. 2016;138:88-99. doi:10.1016/j.neuroimage.2016.05.079
[2] Zhang Z, Ma X,
Dai E, Zhang H, Guo H. Self-calibrated K-space Phase Correction Method for
Multi-shot Diffusion Imaging. In: Proc. Intl. Soc. Mag. Reson. Med. 24
(2016). 2016:3020.
[3] Liao C, Manhard
MK, Bilgic B, et al. Phase-matched virtual coil reconstruction for highly
accelerated diffusion echo-planar imaging. Neuroimage. 2019;194(November
2018):291-302. doi:10.1016/j.neuroimage.2019.04.002
[4] Skare S,
Newbould RD, Clayton DB, Albers GW, Nagle S, Bammer R. Clinical multishot
DW-EPI through parallel imaging with considerations of susceptibility, motion,
and noise. Magn Reson Med. 2007;57(5):881-890. doi:10.1002/mrm.21176
[5] Chen N kuei,
Guidon A, Chang HC, Song AW. A robust multi-shot scan strategy for
high-resolution diffusion weighted MRI enabled by multiplexed
sensitivity-encoding (MUSE). Neuroimage. 2013;72:41-47.
doi:10.1016/j.neuroimage.2013.01.038
[6] Zhang Z, Huang
F, Ma X, Xie S, Guo H. Self-feeding MUSE: A robust method for high resolution
diffusion imaging using interleaved EPI. Neuroimage. 2015;105:552-560.
doi:10.1016/j.neuroimage.2014.10.022
[7] Dietrich O,
Raya JG, Reeder SB, Reiser MF, Schoenberg SO. Measurement of signal-to-noise
ratios in MR images: Influence of multichannel coils, parallel imaging, and
reconstruction filters. J Magn Reson Imaging. 2007;26(2):375-385.
doi:10.1002/jmri.20969