Yu Gao1, Ziwu Zhou1, Fei Han1, Xiaodong Zhong2, Yingli Yang1, and Peng Hu1
1Radiological Sciences, University of California, Los Angeles, Los Angeles, CA, United States, 2MR R&D Collaborations, Siemens Healthineers, Los Angeles, CA, United States
Synopsis
Strong spatial distortion of the DW-ssEPI sequence
prevents its utilization in radiotherapy planning and treatment adaptation. In
this work, a 3D diffusion-prepared magnitude-stabilized bSSFP sequence was
developed and validated at 1.5T. A phase navigator was acquired during the catalyzation
stage of the bSSFP readout to estimate the spatial variation of the signal
phase, and a locally low-rank constrained reconstruction was developed to
resolve the phase variation. The sequence was validated on a diffusion phantom
and healthy volunteers. It provided submillimeter geometric fidelity and acceptable
ADC accuracy, which makes it a promising candidate for treatment planning and adaptation
of the brain.
Introduction
MRI is increasingly used in radiotherapy workflow1. Diffusion-weighted imaging (DWI) has
been shown to have exceptional value for tumor delineation and response
assessment2. A major obstacle that prevents
applying DWI for treatment planning and online adaptive radiotherapy is the
strong spatial distortion in the conventional diffusion-weighted single-shot echo-planar imaging (DW-ssEPI) sequence, which is
problematic for geometrically accurate radiotherapy target delineation. Therefore,
the goal of this work is to develop a 3D diffusion sequence to achieve 1.5mm
isotropic resolution distortion-free diffusion imaging for treatment planning and
radiotherapy treatment adaptation of the brain. Methods
Sequence:
A previously proposed diffusion-prepared
magnitude-stabilized bSSFP (DP-MS-bSSFP, abbreviated as DP-MS) sequence3 was extended to 3D. The sequence
diagram of one example imaging shot is shown in Figure 1. In multi-shot
diffusion imaging, the use of strong diffusion gradients magnifies motion and
system imperfections and creates strong Eddy currents, which then leads to
different signal phase accrual for different k-space segments. A 2D navigator
was acquired during the built-in linear catalyzation pulses to estimate and
correct the MR signal phase inconsistency between different shots. It assumes
that the through-slab phase variation is negligible, and has been shown to be
valid for slab thickness less than 30mm in the brain4,5.
Reconstruction: The multi-shot 3D diffusion data was then phase
corrected slice-by-slice6. A 1D Fourier transform was first applied along
the kz direction. Then for each slice, the k-space data from each individual imaging
shot was used to reconstruct one image. Under ideal scenario where no physiological
motion occurs between shots, the reconstructed image series from multiple shots
should be identical, and hence the rank of the Casorati matrix would be 1.
Therefore, we propose a locally low-rank constrained reconstruction:
$$min_x\sum_{j=1}^{n_c}\sum_{i=1}^L||D_iFS_jP_{ij}x_i-y_{ij} ||_2^2+\gamma \sum_{b\in\Omega} ||C_bx||_\ast$$
where $$$y_{ij}$$$ is the acquired k-space data from the $$$i^{th}$$$ shot and $$$j^{th}$$$ coil, $$$x_i$$$ is the
coil-combined image reconstructed from the $$$i^{th}$$$ shot, $$$x$$$ is the matrix concatenating
all images from different shots, $$$F$$$ is the Fourier
transform, $$$D_i$$$ is the
under-sampling operator for $$$i^{th}$$$ shot, $$$P_{ij}$$$ is the phase
compensation matrix derived from the Gaussian-windowed phase of the built-in
ramp-up navigator, $$$S_j$$$ is the $$$j^{th}$$$ coil
sensitivity map. $$$\Omega$$$ is a set of
small image blocks partitioned from $$$x$$$, $$$C_b$$$ is the operator
that takes image block from the set $$$\Omega$$$ and forms its
Casorati matrix, $$$||\cdot||_{\ast}$$$ is the nuclear
norm, and $$$\gamma $$$ is the regularization
parameter.
Phantom study: A diffusion phantom was scanned on a 1.5T scanner
(Avanto, Siemens Healthineers) to evaluate the geometric fidelity and ADC
accuracy. Scan parameters for the DW-ssEPI sequence and the DP-MS sequence were: TR=6100/2000ms,
TE=86/120ms, field of view=240x240mm2 , resolution=1.5mm3,
bandwidth= 1562/780 Hz/px, average= 12/1, 32 slices (25% oversampling in
DP-MS), scan time=2min46s/9min36s, respectively. b-value of 0 and 800 s/mm2
were acquired. In DP-MS, each kz plane was covered with four shots. The turbo
spin-echo sequence was acquired as the geometric reference. A total of 8
landmarks were selected on each slice of images to calculate the target
registration error (TRE) with the TSE reference, a surrogate of the geometry
accuracy. ADC values from DP-MS were compared with ADCs from DW-ssEPI.
In-vivo study:
Four healthy volunteers were recruited to evaluate the
in-vivo performance. A total of 48 slices in three slabs were acquired to
evaluate and compare the geometric fidelity and ADC. Eight to ten landmarks
were picked in each slice to calculate the TRE. A region of interest (ROI) was
drawn on the white matter for all slices. Additional ROIs were drawn on the
cerebrum and CSF region for slices that contain the cerebrum and the ventricle.
The Bland-Altman analysis was performed to evaluate the ADC agreement between
DW-ssEPI and DP-MS.
Results
As shown Figure 2, DW-ssEPI image had strong
distortion and susceptibility related artifacts, whereas no obvious artifact
was observed in DP-MS image. Compared with TSE images, the maximum landmark
displacement was 8.34mm for DW-ssEPI, and 3.13mm for DP-MS. TREs from DP-MS (0.70±0.50
mm) were significantly lower than TREs from DW-ssEPI (2.09±1.75mm).
Figure 3 shows the phantom ADC results. There was a
good agreement of ADC values measured from the two diffusion sequences. The
percent difference was less than 7%.
An in-vivo comparison is shown in Figure 4. Distortion
and signal pile-up were apparent on DW-ssEPI images, particularly at the base
of the brain. DP-MS provided high-quality diffusion images with no observable
distortion. The mean TRE of DW-ssEPI was over 2mm for all four volunteers, and
the maximum displacement was almost 10mm. The mean TRE was less than 0.8mm for
DP-MS.
The Bland-Altman plots of the mean diffusivity for
selected ROIs on the white matter, cerebellum, and CSF were shown in Figure 5.
Overall, there were good agreements of ADC between the two diffusion
approaches, and the systematic biases were low (0.01x10-3mm2/s, 0.02x10-3mm2/s,
and -0.02x10-3mm2/s for white matter, cerebellum, and
CSF, respectively).Discussion and conclusion
In this work, a 3D diffusion-prepared
magnitude-stabilized bSSFP sequence was developed and validated for
high-quality distortion-free diffusion imaging. It provided submillimeter
geometric fidelity and acceptable ADC accuracy on phantom and volunteer brain
scan, which makes it a promising candidate for treatment planning and adaptive
radiotherapy.Acknowledgements
No acknowledgement found.References
1. Benedict
SH, Yenice KM, Followill D, et al. Stereotactic body radiation therapy: The
report of AAPM Task Group 101. Med Phys. 2010;37(8):4078-4101. doi:10.1118/1.3438081
2. Leibfarth
S, Winter RM, Lyng H, Zips D, Thorwarth D. Potentials and challenges of
diffusion-weighted magnetic resonance imaging in radiotherapy. Clin Transl
Radiat Oncol. 2018;13:29-37. doi:10.1016/j.ctro.2018.09.002
3. Gao
Y, Han F, Zhou Z, et al. Multishot diffusion-prepared magnitude-stabilized
balanced steady-state free precession sequence for distortion-free diffusion
imaging. Magn Reson Med. 2019;81(4):2374-2384. doi:10.1002/mrm.27565
4. Engström
M, Skare S. Diffusion-weighted 3D multislab echo planar imaging for high
signal-to-noise ratio efficiency and isotropic image resolution. Magn Reson
Med. 2013;70(6):1507-1514. doi:10.1002/mrm.24594
5. Chang
H-C, Hui ES, Chiu P-W, Liu X, Chen N-K. Phase correction for three-dimensional
(3D) diffusion-weighted interleaved EPI using 3D multiplexed sensitivity
encoding and reconstruction (3D-MUSER). Magn Reson Med.
2018;79(5):2702-2712. doi:10.1002/mrm.26944
6. Chang
H-C, Sundman M, Petit L, et al. Human brain diffusion tensor imaging at submillimeter
isotropic resolution on a 3Tesla clinical MRI scanner. NeuroImage.
2015;118:667-675. doi:10.1016/j.neuroimage.2015.06.016