Daniel Polak1,2, Daniel Nicolas Splitthoff1, Berkin Bilgic2,3,4, Lawrence L. Wald2,3,4, Kawin Setsompop5, and Stephen F. Cauley2,3,4
1Siemens Healthcare GmbH, Erlangen, Germany, 2Department of Radiology, A. A. Martinos Center for Biomedical Imaging, Charlestown, MA, United States, 3Department of Radiology, Harvard Medical School, Boston, MA, United States, 4Harvard-MIT Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, United States, 5Department of Radiology, Stanford, Stanford, CA, United States
Synopsis
SAMER is a navigation-free retrospective motion-correction
technique which achieves rapid motion estimation using an ultra-fast, low-resolution
scout scan as an image prior. In this work, the SAMER framework is extended to 3D
volumetric reconstructions of 2D TSE imaging data. The optimized 3D volumetric
scout scan is combined with a distributed 2D TSE slice ordering for fully
separable motion estimation with negligible added scan time. The motion
correction performance was evaluated in-vivo for representative motion
trajectories and compatibility to highly accelerated Simultaneous Multi-Slice acquisitions
is demonstrated.
Introduction
Navigator-free retrospective motion correction
techniques [1]–[4] often solve a coupled optimization
problem (Fig. 1) where the data-consistency error of a SENSE+motion
model is minimized with respect to the unknown motion parameters and an image
estimate “jointly”. This poses a computationally demanding non-convex inverse
problem with several hundred temporal motion parameters and millions of imaging
voxels. The recently proposed SAMER technique [5] speeds-up the motion estimation by using
an ultra-fast, low-resolution scout scan as an image prior. This strategy decouples
motion estimation from the image reconstruction and avoids the computationally
demanding repeated updates of an image estimate (Fig. 1). In addition, the
method decouples the motion states themselves leading to a highly scalable
computational solution.
Here, we extend SAMER from 3D volumetric
imaging [5] to 3D volumetric reconstructions of 2D TSE
imaging data (including clinically standard slice gaps). A 3D volumetric SPACE
scout scan was optimized to minimize additional scan time while maintaining contrast.
The 2D TSE slice ordering was improved to increase the temporal resolution of
the motion estimation while maintaining the clinically desired contrast. The
motion correction performance of our technique was evaluated in-vivo for representative
motion trajectories and compatibility to highly accelerated Simultaneous
Multi-Slice (SMS) acquisitions [6] is demonstrated.Methods
Optimization of scout acquisition:
The scout acquisition was optimized to achieve
accurate motion estimation with minimal added scan time. We used a prototype 3D
SPACE sequence (3D TSE) as the scout, as it provides volumetric imaging data without
having to compensate for slice gaps to improve accuracy. To minimize the scan
time of the scout we evaluated the motion estimation robustness when the 3D
SPACE data was acquired at different spatial resolutions and parallel imaging
accelerations. For these analyses, a simulated motion dataset was constructed
from three motion-free 2D TSE scans that contained only intra-scan motion (Fig.
2). Ground truth parameters were obtained from image space registrations. All
scans were acquired on a 3T system (MAGNETOM Vida, Siemens Healthcare,
Erlangen, Germany) using a 32-channel head coil. The 3D volumetric coil
sensitivity maps were computed from an optimized 2 sec external GRE reference scan
using ESPIRiT [7].
Optimization of 2D TSE slice ordering:
The 2D TSE slice ordering was optimized to
allow for higher temporal estimation of the motion. Standard T2w TSE
acquisitions often use long TRs (~6 sec) which is insufficient to capture
typical patient motion. Higher temporal resolution can be achieved by intra-TR
data binning. However, the smaller amount of slice-ky data used can
lead to decreased motion estimation accuracy. We introduce an interleaved slice
ordering scheme with complementary ky sampling and compare the robustness
of the motion estimation to a standard sampling pattern. The simulated motion
data described above (Fig. 2) was used and intra-TR data binning was performed
by grouping the data from each TR into three motion bins.
In vivo motion experiments:
The motion correction performance of our
approach was evaluated in-vivo for axial T2w TSE with supervised subject motion
(in-plane and/or through-plane). In addition, the compatibility to highly
accelerated Simultaneous Multi-Slice acquisitions (R=2, MB=2) with CAIPIRINHA
sampling [8] was demonstrated. In all motion
experiments, a R=12-fold accelerated 3D SPACE scout (TA=6 sec, Fig. 2) and an
interleaved 2D TSE ordering scheme with complementary ky sampling (Fig.
3) were used to estimate 3D motion every two seconds (three motion bins per TR).Results
Accurate estimation
of 3D motion within 2D TSE imaging data was achieved using an optimized, six-second
3D volumetric SPACE scout acquired at low spatial resolution and high
acceleration (Fig. 2).
Intra-TR
data binning with standard ascending slice ordering yielded inaccurate motion
estimation, as opposed to an interleaved slice ordering with complementary ky
sampling (Fig. 3).
Reduction of motion artifacts was achieved for
all in-vivo acquisitions (Fig. 4+5). SAMER improved delineation of fine
anatomical structures, e.g., vessels (yellow arrows), and resulted in a
decrease of the data-consistency-error (DC). However, despite a full 3D
reconstruction, residual motion artifacts and some loss of spatial resolution (red
arrows) were observed, especially, in the case of through-plane rotation.Discussion
In this work, we demonstrated 3D motion artifact
mitigation for standard clinical 2D TSE imaging where the use of a rapid 6 sec 3D
volumetric scout enabled highly efficient and separable motion estimation and
correction. We evaluated the motion mitigation performance in-vivo and observed
robust artifact reduction across a variety of motion trajectories.
Intra-TR data binning was used to increase the
temporal resolution of the 3D motion estimation. Here, three bins per TR were
used which led to a three-fold improvement in temporal resolution but with
proportional increases to the reconstruction. Dynamically binning the data is
likely to improve the trade-off of motion correction accuracy and
reconstruction time.
Through-plane interpolation across thick
slices, spin history, and slice gaps caused residual motion artifacts despite the
use of a full 3D reconstruction model. Future work will investigate the use of
data/slice over-sampling [4] and/or machine learning [10] to compensate for missing or
degraded data. Moreover, SMS can allow for a flexible trade-off between
additional scan time and image quality.Acknowledgements
No acknowledgement found.References
[1] M. W. Haskell, S. F. Cauley, and L. L.
Wald, “TArgeted Motion Estimation and Reduction (TAMER): Data consistency based
motion mitigation for mri using a reduced model joint optimization,” IEEE
Trans. Med. Imaging, vol. 37, no. 5, pp. 1253–1265, 2018.
[2] A.
Loktyushin, H. Nickisch, R. Pohmann, and B. Schölkopf, “Blind multirigid
retrospective motion correction of MR images,” Magn. Reson. Med., vol.
73, no. 4, pp. 1457–1468, 2015.
[3] L.
Cordero-Grande, R. P. A. G. Teixeira, E. J. Hughes, J. Hutter, A. N. Price, and
J. V. Hajnal, “Sensitivity Encoding for Aligned Multishot Magnetic Resonance
Reconstruction,” IEEE Trans. Comput. Imaging, vol. 2, no. 3, pp.
266–280, 2016.
[4] L.
Cordero-Grande, E. J. Hughes, J. Hutter, A. N. Price, and J. V. Hajnal,
“Three-dimensional motion corrected sensitivity encoding reconstruction for
multi-shot multi-slice MRI: Application to neonatal brain imaging,” Magn.
Reson. Med., vol. 79, no. 3, pp. 1365–1376, 2018.
[5] D.
Polak, S. Cauley, D. Splitthoff, P. Bachert, and K. Setsompop, “Scout
Acquisition enables rapid Motion Estimation (SAME) for fully separable
retrospective motion mitigation,” in Proceedings of ISMRM 28th Annual
Meeting, 2020.
[6] D. J.
Larkman, J. V. Hajnal, A. H. Herlihy, G. A. Coutts, I. R. Young, and G.
Ehnholm, “Use of multicoil arrays for separation of signal from multiple slices
simultaneously excited,” J. Magn. Reson. Imaging, 2001.
[7] M.
Uecker et al., “ESPIRiT - An eigenvalue approach to autocalibrating
parallel MRI: Where SENSE meets GRAPPA,” Magn. Reson. Med., vol. 71, no.
3, pp. 990–1001, Mar. 2014.
[8] F. A.
Breuer, M. Blaimer, R. M. Heidemann, M. F. Mueller, M. A. Griswold, and P. M.
Jakob, “Controlled aliasing in parallel imaging results in higher acceleration
(CAIPIRINHA) for multi-slice imaging,” Magn. Reson. Med., vol. 53, no.
3, pp. 684–691, 2005.
[9] D.
Polak et al., “Joint multi-contrast variational network reconstruction
(jVN) with application to rapid 2D and 3D imaging,” Magn. Reson. Med.,
2020.
[10] M. W.
Haskell et al., “Network Accelerated Motion Estimation and Reduction
(NAMER): Convolutional neural network guided retrospective motion correction
using a separable motion model,” Magn. Reson. Med., vol. 82, no. 4, pp. 1452–1461, 2019.