Azusa Sakurama1, Yasutaka Fushimi1, Satoshi Nakajima1, Yusuke Yokota1, Sonoko Oshima1, Sayo Otani1, Krishna Pandu Wicaksono1, Tomohisa Okada2, Wei Liu3, and Kaori Togashi1
1Department of Diagnostic Imaging and Nuclear Medicine, Kyoto University Graduate School of Medicine, Kyoto, Japan, 2Human Brain Research Center, Kyoto University Graduate School of Medicine, Kyoto, Japan, 3Siemens Healthineers AG, Shenzhen, China
Synopsis
This comparison study between MPRAGE
with GRAPPA and MPRAGE with Wave-CAIPI has been performed at the clinical MR
scanner. MPRAGE with Wave-CAIPI 3×3 shows relatively good contrast despite its
short scan time of 1 m 42 s. No segmentation error was found in MPRAGE with
Wave-CAIPI 3×3 in our cohort. The correlation of cerebral cortex and cerebral
white matter was very high between 2 MPRAGEs. In deep gray matters except
pallidum and hippocampus, the correlation of VOIs were also high. These results suggest that volumes derived from
automated segmentation of MPRAGE with Wave-CAIPI are reliable measures.
Introduction
By
taking advantage of coil sensitivity encoding from multi-channel receiver
coils, reduction of phase-encoding steps has been available with the parallel
imaging technique such as SENSE and GRAPPA. Controlled aliasing in parallel
imaging results in higher acceleration (2D-CAIPIRINHA) has enabled image
reconstruction with further acceleration, which can efficiently employ the
sensitivity variation in phase encoding directions [1]. To further utilize the sensitivity
information in the readout direction, bunch phase encoding (BPE) has been combined
with parallel imaging to improve the image reconstruction [2-3]. Wave-CAIPI
acquisition extends and combines the BPE and 2D-CAIPIRINHA by playing
sinusoidal gradients during the readout of each phase encoding line and introducing
interslice shifts across the aliasing slices [4-5], which enables highly
accelerated volumetric imaging with low artifact and signal-to-noise ratio
(SNR) penalties. More acceleration has been expected with the introduction of
Wave-CAIPI, however, image quality has not been evaluated well in the clinical
practices.
The purpose of this study is to evaluate MPRAGE with
9-fold acceleration by using Wave-CAIPI (Wave-CAIPI 3×3) in clinical scanners,
and to compare MPRAGE Wave-CAIPI 3×3 with conventional MPRAGE GRAPPA 2×1.Method
Healthy
volunteers and patients with neurological disorders were enrolled in this
study. 11 patients were excluded due to severe head motion, cerebral infarction
and hemorrhage, and the total number of 50 subjects were included in this
study. This study was approved by the institutional review board and written
informed consent was obtained from each subject.
Subjects
underwent MR imaging at 3T MR scanners (MAGNETOM Prisma, Siemens Healthineers,
Erlangen, Germany) with a 64-channel head/neck coil.
The
imaging protocols of two MPRAGEs were as follows: (i) MPRAGE with GRAPPA 2×1,
ref lines PE, 24; 208 slices; scan time, 5 m 21 s. (ii) a prototype MPRAGE with
Wave-CAIPI 3×3; ref lines PE, 24; ref lines 3D, 24; CAIPI shift factor, 1; 240
slices; scan time, 1 m 42 s. TR of 2300 ms, TE of 4.67ms, TI of 922ms, flip
angle of 9, and bandwidth of 130 Hz/pixel were applied to both MPRAGEs.
Recon-all
function of FreeSurfer (v6.0.1)
(http://surfer.nmr.mgh.harvard.edu/)
was applied for both MPRAGE images [6]. The Parcellation status was visually
checked and voxel sizes of VOIs were compared.
Results
No
segmentation error was found in both MPRAGEs in our cohort.
Representative
images of original MPRAGE and segmented images by using recon-all function of
FreeSurfer were shown in Figure 1 and 2. MPRAGE GRAPPA 2×1 showed better image
quality, however, MPRAGE Wave-CAIPI 3×3 shows relatively good contrast between
gray and white matters in healthy subjects (Figure 1). Of note, MPRAGE Wave-CAIPI
3×3 also showed nice segmented results even in patients with brain atrophy
(Figure 2).
We
compared the volumes of each VOIs between GRAPPA 2×1 and Wave-CAIPI 3×3.
The correlation
of all VOIs was 0.991 [0.989 - 0.992] (Figure 3). Volumes of each VOI and the
correlation between GRAPPA 2×1 and Wave-CAIPI 3×3 are shown in Figure 4 and 5. The
results of cerebral white matter and cerebral cortex were shown in Figure 4.
The
correlation of cerebral cortex and cerebral white matter was very high as
0.98-0.99.
The
results of deep gray matters, hippocampus and amygdala were shown in Figure 5.
The correlation of almost VOIs were higher than 0.9.
Discussion
Wave-CAIPI
MPRAGE was evaluated in comparison with other accelerated techniques of MPRAGE.
MPRAGE with Wave-CAIPI 3×3 shows relatively better contrast despite of its
short scan time. Less imaging noise and segmentation error was found in MPRAGE
with Wave-CAIPI 3×3.
Wave-CAIPI showed good image quality for its high
acceleration factor. Short scan time of Wave-CAIPI can reduce motion artifact
and high SNR derived from 64-channel coil may attribute to relatively good image
quality.
The correlation
of cerebral cortex and cerebral white matter was very high as 0.98-0.99.
From
our result, we can completely substitute MPRAGE GRAPPA 2×1 with MPRAGE
Wave-CAIPI 3×3 for volumetric evaluation of cerebral white matter and cerebral
cortex.
In deep
gray matters except pallidum and hippocampus, the correlation of VOIs were
higher than 0.9. The correlation of pallidum and amygdala were relatively lower
than that of other VOIs. This is probably because pallidum and amygdala are relatively
small in volume. In addition, pallidum and amygdala show less image contrast on
both MPRAGE GRAPPA 2×1 and Wave-CAIPI 3×3, therefore, parcellation of these
VOIs were derived from probabilistic image segmentation.
The
structural volumes of MPRAGE Wave-CAIPI 3×3 were almost similar to MPRAGE
GRAPPA 2×1 in average. SNR should be lower in Wave-CAIPI 3×3,
compared with Grappa 2×1. However, based on our results of FreeSurfer, putative
lower SNR of Wave-CAIPI seems to have no impact on segmentation.
Conclusion
This
comparison study between MPRAGE with GRAPPA and MPRAGE with Wave-CAIPI has been
performed at the clinical MR scanner. MPRAGE with Wave-CAIPI 3×3 shows
relatively better contrast despite of its short scan time of 1 m 42 s. No
segmentation error was found in MPRAGE with Wave-CAIPI 3×3. The correlation of
cerebral cortex and cerebral white matter was very high. In deep gray matters
except pallidum and hippocampus, the correlation of VOIs were high. These results suggest that
volumes derived from automated segmentation of MPRAGE with Wave-CAIPI are
reliable measures. Acknowledgements
We are grateful to Mr. Yuta Urushibata, Siemens
Healthcare K. K., for helpful discussions.References
-
Breuer
F, Blaimer M, Mueller MF, Seiberlich N, Heidemann RM,
Griswold MA, Jakob PM. Controlled aliasing in
volumetric parallel
imaging (2D CAIPIRINHA).
Magn Reson Med 2006;55.3:549–556.
- Moriguchi H, Duerk JL. Magn Reson Med. 2006
Mar;55(3):633-48. Bunched phase encoding (BPE): a new fast data acquisition
method in MRI.
- Breuer FA, Moriguchi H, Seiberlich
N, Blaimer M, Jakob PM, Duerk JL, Griswold MA. Magn Reson Med. 2008 Aug;60(2):474-8. Zigzag sampling for
improved parallel imaging.
- Bilgic B, Gagoski BA, Cauley SF, Fan AP, Polimeni
JR, Grant PE, Wald LL, Setsompop K. Magn Reson Med. 2015 Jun;73(6):2152-62.
Wave-Caipi for highly accelerated 3D imaging.
- Daniel P., Kawin S., Stephen F. Magn Reson Med.
2018 Jan; 79(1): 401–406. Wave-CAIPI for Highly Accelerated MP-RAGE Imaging.
- Dale, A.M., Fischl, Bruce, Sereno, M.I. Neuroimage.
1999 Feb;9(2):179-94. Cortical Surface-Based Analysis I: Segmentation and
Surface Reconstruction.