Shihui Chen1, Liyuan Liang1,2, Chenglang Yuan1, and Hing-Chiu Chang1,2
1The Department of Biomedical Engineering, The Chinese University of Hong Kong, Hong Kong, Hong Kong, 2Multi-scale Medical Robotics Center, Hong Kong, Hong Kong
Synopsis
Keywords: Diffusion Reconstruction, Diffusion Tensor Imaging, 2D navigator, eddy current, phase correction, High Angular Resolution Diffusion Imaging
Motivation: The time-varying eddy current caused by diffusion gradient can induce additional phase difference between imaging and navigator echoes. These inaccurate measurements may cause residual artifacts in the subsequently reconstructed data.
Goal(s): This study aims to verify the presence of the eddy-current-induced phase differences in navigator echoes and the phase correction can improve reconstruction performance on in-vivo DTI and HARDI.
Approach: We proposed a procedure to calibrate eddy-current-induced phase difference from a phantom and considered the additional phase difference when using k-d SVD method for reconstruction.
Results: The robust reconstruction performance can be achieved by using phase-corrected navigator data to recover highly-undersampled data.
Impact: This study demonstrates that the eddy-current-induced
phase differences between imaging and navigator echoes can be calibrated in
advance, and then corrected during the k-d SVD reconstruction method to enable
highly-accelerated multi-shot high angular resolution diffusion imaging (HARDI).
Introduction
The successful reconstruction of
multi-shot diffusion tensor imaging (DTI) relies on proper correction of
inter-shot phase variations, and 2D navigator echo is commonly used in multi-shot
DTI for measuring this type of data inconsistencies(1,2). However, the time-varying eddy
current associated with diffusion gradient(3) can induce additional phase difference
between imaging and navigator echoes, leading to inaccurate measurement of
inter-shot phase variations that may cause residual artifacts in the
reconstructed diffusion-weighted images. In this study, we first verified the presence of
eddy-current-induced phase difference between imaging and navigator echoes and
its dependency on diffusion direction. Then, we proposed a procedure to calibrate
the eddy-current-induced phase difference from a phantom for each diffusion
direction. Finally, the k-d SVD reconstruction method(4) was further modified to take into
account the pre-calibrated phase differences for reconstructing
highly-accelerated multi-shot high angular resolution diffusion imaging (HARDI)
data without any residual artifacts. Methods
Design of pulse sequence
The 2D navigator echo was implemented
into an interleaved diffusion-weighted EPI sequence (Fig.1a). For investigating
the presence of eddy-current-induced phase differences between imaging and
navigator echoes, the Gro and Gpe gradients (grey-shaded in Fig.1a) were turned off and
only the FID signal was acquired for each EPI readout.
Data acquisition
Three sessions of data were
collected on a 1.5T GE MRI scanner using the proposed
pulse sequence (Fig.1a) with a 12-channel head coil. In Session 1, a series of FID
signals between imaging and navigator echoes at different diffusion directions
were obtained by turning off the Gro and Gpe gradients. In
Sessions 2&3, two HARDI datasets were respectively collected from a phantom
and a healthy volunteer with the identical scan parameters as follows: matrix
size=128$$$\times$$$128, FOV=24cm, b-value=0/800 s/mm2,
number of shots=4, 64 diffusion directions with a spherical ordering(5). Four-fold undersampled dataset was
simulated by selecting one of k-space segments for each diffusion direction in
a cyclical manner (Fig.1b).
Data reconstruction and analysis
For a series of acquired FID
signals (Session 1), the phase change slope of each FID was derived after
applying 1D phase unwrapping. Then all slopes at different time points were
compared between imaging and navigator echoes to verify the presence of time-varying
eddy current. For the phantom data (Session 2), different k-space segments were
individually reconstructed with SENSE(6) for both imaging and navigator
echoes. The baseline phase was removed by applying
the coil sensitivity profiles derived from imaging and navigator data at b = 0
s/mm2 respectively. The phase difference between imaging and
navigator echoes at each specific diffusion direction was fitted with a
first-order 2D polynomial function. For the in-vivo brain data (Session 3), the
fitted phase difference maps were used to remove the additional phase
variations due to eddy current effect for all navigator echoes. Afterward, the
navigator echoes of different diffusion directions with corrected phase were
used as the training data for k-d SVD reconstruction(4), and recovered the
highly-undersampled HARDI data (Fig.2). For comparison, the imaging data was
reconstructed using MUSE(7) with the inter-shot phase
variations measured from either imaging or navigator echoes. Results
Fig.3 shows the comparison of phase
change slopes across time in imaging and navigator echo at b = 0/800 s/mm2 with two different diffusion directions. Fig.4 shows
the phase difference maps acquired from phantom and the corresponding fitted
maps across eight different diffusion directions. Fig.4 also shows the phases of
navigator echoes with and without phase correction, and the true phase maps measured
from imaging echo in human brain for comparison. Fig.5 shows the reconstructed diffusion-weighted images of eight diffusion directions using MUSE and our proposed method with
and without applying phase correction in navigator data. Discussion
Our study has verified that the presence
of phase difference between imaging and navigator echoes were caused by the
time-varying eddy current of diffusion gradients. The eddy-current-induced
phase variations can be calibrated from a phantom in advance, and no
extra calibration scan is required for in-vivo HARDI application with k-d SVD
method. For conventional MUSE with linear reconstruction model, the residual
phase difference between imaging and navigator echoes only caused incomplete
phase correction on the real image data. However, the proposed k-d SVD method
required a training data acquired from navigator echoes with reliable phase
information to recover image from highly-undersampled data. Therefore, the
correction for phase difference between imaging and navigator echo is a crucial
step for achieving robust reconstruction performance. Conclusion
Our study demonstrates the presence
of eddy-current-induced phase difference between imaging and navigator echoes,
and applying the phase correction for navigator data can improve the
reconstruction performance of highly-accelerated HARDI data using k-d SVD
method. Acknowledgements
The work was in part supported by grants from Hong Kong
Research Grant Council (GRF17106820, GRF17125321, GRF14206723, and
ECS24213522).References
1. Jeong HK, Gore JC, Anderson AW. High‐resolution human diffusion
tensor imaging using 2‐D
navigated multishot SENSE EPI at 7 T. Magnetic resonance in medicine
2013;69(3):793-802.
2.Jeong
HK, Dewey BE, Hirtle JA, Lavin P, Sriram S, Pawate S, Gore JC, Anderson AW,
Kang H, Smith SA. Improved diffusion tensor imaging of the optic nerve using
multishot two‐dimensional
navigated acquisitions. Magnetic resonance in medicine 2015;74(4):953-963.
3.Truong
T-K, Chen N-k, Song AW. Dynamic correction of artifacts due to susceptibility
effects and time-varying eddy currents in diffusion tensor imaging. NeuroImage
2011;57(4):1343-1347.
4.Chen
S, Chang H-C. Highly-Accelerated Multi-shot Diffusion Imaging with High Angular
Resolution Enabled by kd SVD.
5.Chao
TC, Chiou JyG, Maier SE, Madore B. Fast diffusion imaging with high angular
resolution. Magnetic resonance in medicine 2017;77(2):696-706.
6.Pruessmann
KP, Weiger M, Scheidegger MB, Boesiger P. SENSE: sensitivity encoding for fast MRI.
Magnetic resonance in medicine 1999;42(5):952-962.
7.Chen N-k, Guidon A, Chang H-C, 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.