pith. sign in

arxiv: 1906.10178 · v1 · pith:OVCBQ2M5new · submitted 2019-06-24 · ⚛️ physics.med-ph · eess.IV

Improved Reconstruction for high-resolution Multi-shot Diffusion Weighted Imaging

Merry Mani , Hemant Kumar Aggarwal , Vincent Magnotta , Mathews Jacob This is my paper

Pith reviewed 2026-05-25 16:31 UTC · model grok-4.3

classification ⚛️ physics.med-ph eess.IV
keywords multi-shot diffusion weighted imagingiterative reweighted least squaresconjugate symmetry priorparallel imaging reconstructionhigh-resolution DWIpartial Fourier acquisitioninter-shot motion correctionMUSSELS
0
0 comments X

The pith

An iterative reweighted least squares formulation speeds multi-shot diffusion MRI reconstruction by a factor of six while supporting conjugate symmetry priors that reduce blurring.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes replacing the original MUSSELS reconstruction with an iterative reweighted least squares solver for multi-shot diffusion weighted imaging. This change reduces computation time roughly sixfold on 192 by 192 and 256 by 256 matrices. The new solver also accepts an additional conjugate symmetry constraint on k-space data, which cuts blurring and retains fine detail from partial Fourier acquisitions. If the claims hold, high-resolution whole-brain diffusion studies become practical on standard hardware without specialized phase compensation steps.

Core claim

The IRLS formulation of the MUSSELS parallel imaging reconstruction recovers artifact-free diffusion weighted images from multi-shot EPI data without phase compensation, runs approximately six times faster than prior implementations for typical high-resolution matrix sizes, and accommodates conjugate symmetry priors that reduce blurring while preserving detail in partial Fourier data.

What carries the argument

The iterative reweighted least squares (IRLS) solver applied to the MUSSELS parallel imaging objective, extended to enforce conjugate symmetry on the k-space data.

If this is right

  • Reconstruction time drops by a factor of about six for 192x192 and 256x256 matrices compared with earlier MUSSELS code.
  • Conjugate symmetry enforcement reduces blurring and maintains high-resolution features from partial Fourier sampling.
  • Overall computational cost becomes comparable to conventional multi-shot DWI pipelines.
  • Routine whole-brain high-resolution diffusion studies become feasible with minimal added burden.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same IRLS structure could be reused for other parallel imaging problems that currently scale poorly with matrix size.
  • Further priors beyond conjugate symmetry might be added without changing the core solver loop.
  • Clinical workflows could shift to higher-resolution protocols without extending scan duration or requiring dedicated reconstruction servers.

Load-bearing premise

Adding the conjugate symmetry prior does not create new artifacts or degrade the inter-shot motion correction that MUSSELS was built to perform.

What would settle it

A side-by-side test on motion-corrupted multi-shot datasets showing that reconstructions using the conjugate symmetry prior contain more residual artifacts or greater signal loss than the original MUSSELS method.

Figures

Figures reproduced from arXiv: 1906.10178 by Hemant Kumar Aggarwal, Mathews Jacob, Merry Mani, Vincent Magnotta.

Figure 1
Figure 1. Figure 1: Illustration of the direct reconstruction of msDW [PITH_FULL_IMAGE:figures/full_fig_p016_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Each column wj of W is a multi-channel annihilation filter. The term H1(mb )W effectively computes a multi-channel convolution equivalent to the computation shown on the right side. Here, the k-space matrices mˆ are convolved with several multi-channel filters w of size r × r × Ns. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison of SVS MUSSELS reconstruction (a) and IRLS MUSSELS re￾construction (b). While formulation gives equivalent results, the IRLS MUSSELS re￾construction is much faster compared to the SVS MUSSELS reconstruction. (c) & (d) shows the color coded directional FA map computed from all the DWIs corresponding to the two reconstructions. 18 [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: DWI reconstruction from various slice locations s [PITH_FULL_IMAGE:figures/full_fig_p019_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FA maps reconstructed from several slice location [PITH_FULL_IMAGE:figures/full_fig_p020_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Color-coded FA maps reconstructed from several sl [PITH_FULL_IMAGE:figures/full_fig_p021_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of IRLS MUSSELS reconstruction without (a) and with (b) CS performed on dataset 2. (b) shows sharper recov￾ery of the data and the anatomical details are better defined than in (a) as indicated by the arrows. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Whole brain reconstruction of dataset 2. The IRLS w [PITH_FULL_IMAGE:figures/full_fig_p024_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Whole brain reconstruction from dataset 2. The ODF [PITH_FULL_IMAGE:figures/full_fig_p025_9.png] view at source ↗
Figure 1
Figure 1. Figure 1: Illustration of the direct reconstruction of msDW dat [PITH_FULL_IMAGE:figures/full_fig_p026_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Each column wj of W is a multi-channel annihilation filter. The term H1(mb )W effectively com￾putes a multi-channel convolution equivalent to the computation shown on the right side. Here, the k-space matrices mˆ are convolved with several multi-channel filters w of size r × r × Ns [PITH_FULL_IMAGE:figures/full_fig_p026_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: DWI reconstruction from various slice locations show [PITH_FULL_IMAGE:figures/full_fig_p026_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FA maps reconstructed from several slice locations sh [PITH_FULL_IMAGE:figures/full_fig_p026_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Color-coded FA maps reconstructed from several slice [PITH_FULL_IMAGE:figures/full_fig_p026_6.png] view at source ↗
Figure 9
Figure 9. Figure 9: Whole brain reconstruction from dataset 2. The ODFs re [PITH_FULL_IMAGE:figures/full_fig_p027_9.png] view at source ↗
read the original abstract

Purpose: To introduce a fast and improved direct reconstruction method for multi-shot diffusion weighted (msDW) scans for high-resolution studies. Methods:Multi-shot EPI methods can enable higher spatial resolution for diffusion MRI studies. Traditionally, such acquisitions required specialized reconstructions involving phase compensation to correct for inter-shot motion artifacts. The recently proposed MUSSELS reconstruction belongs to a new class of parallel imaging-based methods that recover artifact-free DWIs from msDW data without needing phase compensation. However, computational demands of the MUSSELS reconstruction scales as the matrix size and the number of shots increases, which hinders its practical utility for high-resolution applications. In this work, we propose a computationally efficient formulation using iterative reweighted least squares (IRLS) method. The new formulation is not only fast but it enables to accommodate additional priors such as conjugate symmetry property of the k-space data to improve the reconstruction. Using whole-brain in-vivo data, we show the utility of the new formulation for routine high-resolution studies with minimal computational burden. Results: The IRLS formulation provides about six times faster reconstruction for matrix sizes 192x192 and 256x256, compared to the original implementations. The reconstruction quality is improved by the addition of conjugate symmetry priors that reduce blurring and preserves the high-resolution details from partial Fourier acquisitions. Conclusion: The proposed method is shown to be computationally efficient to enable routine high-resolution studies. The computational complexity matches the traditional msDWI reconstruction methods and provides improved reconstruction results.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The manuscript proposes an iterative reweighted least squares (IRLS) reformulation of the MUSSELS reconstruction for multi-shot diffusion-weighted EPI. It claims this yields an approximately six-fold reduction in reconstruction time for 192×192 and 256×256 matrices relative to prior MUSSELS implementations, while the addition of a conjugate-symmetry (Hermitian) prior on k-space data reduces blurring and better preserves high-resolution detail from partial-Fourier acquisitions, all without explicit inter-shot phase compensation.

Significance. If the speed-up and quality claims are substantiated with quantitative metrics and motion-robustness validation, the work would make routine high-resolution msDWI clinically feasible by bringing reconstruction cost in line with conventional parallel-imaging methods while exploiting partial-Fourier data. The absence of error metrics, statistical tests, or explicit checks on the interaction between the new prior and MUSSELS motion handling currently prevents a firm assessment of practical impact.

major comments (3)
  1. [Abstract] Abstract (Results paragraph): The statement that the IRLS formulation “provides about six times faster reconstruction” for the cited matrix sizes supplies no timing tables, hardware specifications, shot counts, or direct comparison against the original MUSSELS solver; without these data the quantitative speed-up claim cannot be evaluated.
  2. [Abstract] Abstract (Results paragraph): The claim that conjugate-symmetry priors “reduce blurring and preserve the high-resolution details” and are “accommodated” without degrading inter-shot motion correction is unsupported by any residual-phase maps, motion-parameter error statistics, or image-quality metrics on data containing known inter-shot phase errors. This interaction is load-bearing for the central claim that the method retains MUSSELS motion robustness.
  3. [Abstract] Abstract (Methods/Results): No description is given of how the IRLS solution was validated against the original MUSSELS implementation (e.g., normalized root-mean-square error, structural similarity, or difference images), leaving the assertion of “improved reconstruction results” without visible quantitative grounding.
minor comments (1)
  1. [Abstract] The abstract repeatedly uses “accommodated” and “preserves” without defining the precise mathematical incorporation of the Hermitian prior into the IRLS objective; a short methods paragraph clarifying the augmented cost function would improve clarity.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive comments. We address each major comment point by point below, indicating where revisions will be made to the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract (Results paragraph): The statement that the IRLS formulation “provides about six times faster reconstruction” for the cited matrix sizes supplies no timing tables, hardware specifications, shot counts, or direct comparison against the original MUSSELS solver; without these data the quantitative speed-up claim cannot be evaluated.

    Authors: We agree that the abstract would benefit from additional specifics. The full manuscript reports timing comparisons for the cited matrix sizes in the Results section. We will revise the abstract to reference the hardware platform, shot counts, and direct comparison details, and will ensure a timing table is clearly presented in the main text. revision: yes

  2. Referee: [Abstract] Abstract (Results paragraph): The claim that conjugate-symmetry priors “reduce blurring and preserve the high-resolution details” and are “accommodated” without degrading inter-shot motion correction is unsupported by any residual-phase maps, motion-parameter error statistics, or image-quality metrics on data containing known inter-shot phase errors. This interaction is load-bearing for the central claim that the method retains MUSSELS motion robustness.

    Authors: The in-vivo whole-brain experiments demonstrate reduced blurring from the conjugate-symmetry prior on partial-Fourier data while retaining the motion-robust property of MUSSELS. To strengthen the claim, we will add explicit supporting material such as residual phase maps or image-quality metrics comparing reconstructions with and without the prior on data with inter-shot phase errors. revision: yes

  3. Referee: [Abstract] Abstract (Methods/Results): No description is given of how the IRLS solution was validated against the original MUSSELS implementation (e.g., normalized root-mean-square error, structural similarity, or difference images), leaving the assertion of “improved reconstruction results” without visible quantitative grounding.

    Authors: We will revise the manuscript to include a description of the validation procedure, incorporating quantitative metrics such as normalized root-mean-square error and difference images between the IRLS and original MUSSELS reconstructions on the in-vivo datasets. revision: yes

Circularity Check

0 steps flagged

No significant circularity; reformulation and prior are independent of fitted outputs

full rationale

The paper introduces an IRLS reformulation of the existing MUSSELS method and adds the standard conjugate-symmetry k-space prior; neither step is defined in terms of its own outputs, fitted parameters, or a self-citation chain. All reported speed-ups and quality gains are measured on external in-vivo acquisitions rather than recovered by construction from the same data used to tune the algorithm. No load-bearing uniqueness theorem or ansatz is smuggled in via prior self-work, and the derivation remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard MRI parallel-imaging assumptions plus the domain assumption that conjugate symmetry holds sufficiently well in the acquired k-space data to be used as a prior.

axioms (1)
  • domain assumption Conjugate symmetry property of k-space data holds for the partial Fourier acquisitions
    Invoked in the abstract to improve reconstruction quality from partial Fourier data.

pith-pipeline@v0.9.0 · 5804 in / 1199 out tokens · 51224 ms · 2026-05-25T16:31:12.939835+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

  1. [1]

    Image formation in diffusion MRI: A review of recent technical developments

    Wu W and Miller KL. Image formation in diffusion MRI: A review of recent technical developments . Journal of Magnetic Resonance Imaging , 46 0 (3): 0 646--662, 2017

    work page 2017

  2. [2]

    Quantifying brain microstructure with diffusion MRI: Theory and parameter estimation

    Novikov DS, Fieremans E, Jespersen SN, and Kiselev VG. Quantifying brain microstructure with diffusion MRI: Theory and parameter estimation . NMR in Biomedicine , page 3998, 2018

    work page 2018

  3. [3]

    Microstructural imaging of the human brain with a ‘super-scanner': 10 key advantages of ultra-strong gradients for diffusion MRI

    Jones D, Alexander D, Bowtell R, Cercignani M, Dell'Acqua F, McHugh D, Miller K, Palombo M, Parker G, Rudrapatna U, and Tax C. Microstructural imaging of the human brain with a ‘super-scanner': 10 key advantages of ultra-strong gradients for diffusion MRI . NeuroImage , 182: 0 8--38, 2018

    work page 2018

  4. [4]

    Analysis and correction of motion artifacts in diffusion weighted imaging

    Anderson AW and Gore JC. Analysis and correction of motion artifacts in diffusion weighted imaging. Magnetic Resonance in Medicine , 32 0 (3): 0 379--87, 1994

    work page 1994

  5. [5]

    Real diffusion-weighted MRI enabling true signal averaging and increased diffusion contrast

    Eichner C, Cauley SF, Cohen-Adad J, M \" o ller HE, Turner R, Setsompop K, and Wald LL. Real diffusion-weighted MRI enabling true signal averaging and increased diffusion contrast . NeuroImage , 122: 0 373--384, 2015

    work page 2015

  6. [6]

    Diffusion-weighted interleaved echo-planar imaging with a pair of orthogonal navigator echoes

    Butts K, de Crespigny A, Pauly JM, and Moseley M. Diffusion-weighted interleaved echo-planar imaging with a pair of orthogonal navigator echoes. Magnetic Resonance in Medicine , 35 0 (5): 0 763--70, 1996

    work page 1996

  7. [7]

    A robust multi-shot scan strategy for high-resolution diffusion weighted MRI enabled by multiplexed sensitivity-encoding (MUSE)

    Chen NK, Guidon A, Chang HC, and Song AW. A robust multi-shot scan strategy for high-resolution diffusion weighted MRI enabled by multiplexed sensitivity-encoding (MUSE) . NeuroImage , 2013

    work page 2013

  8. [8]

    Multi-shot sensitivity-encoded diffusion data recovery using structured low-rank matrix completion (MUSSELS)

    Mani M, Jacob M, Kelley D, and Magnotta V. Multi-shot sensitivity-encoded diffusion data recovery using structured low-rank matrix completion (MUSSELS) . Magnetic Resonance in Medicine , 78 0 (2): 0 494--507, 2017

    work page 2017

  9. [9]

    Comprehensive reconstruction of multi-shot multi-channel diffusion data using mussels

    Mani M, Magnotta V, Kelley D, and Jacob M. Comprehensive reconstruction of multi-shot multi-channel diffusion data using mussels . In Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS , 2016. ISBN 9781457702204

    work page 2016

  10. [10]

    Motion-robust reconstruction of multishot diffusion-weighted images without phase estimation through locally low-rank regularization

    Hu Y, Levine EG, Tian Q, Moran CJ, Wang X, Taviani V, Vasanawala SS, McNab JA, Daniel BA, and Hargreaves BL. Motion-robust reconstruction of multishot diffusion-weighted images without phase estimation through locally low-rank regularization . Magnetic Resonance in Medicine , 2018

    work page 2018

  11. [11]

    Readout-segmented EPI for rapid high resolution diffusion imaging at 3T

    Holdsworth SJ, Skare S, Newbould RD, Guzmann R, Blevins NH, and Bammer R. Readout-segmented EPI for rapid high resolution diffusion imaging at 3T . European Journal of Radiology , 2008

    work page 2008

  12. [12]

    High resolution diffusion-weighted imaging using readout-segmented echo-planar imaging, parallel imaging and a two-dimensional navigator-based reacquisition

    Porter DA and Heidemann RM. High resolution diffusion-weighted imaging using readout-segmented echo-planar imaging, parallel imaging and a two-dimensional navigator-based reacquisition . Magnetic Resonance in Medicine , 62 0 (2): 0 468--475, 2009

    work page 2009

  13. [13]

    Self-navigated interleaved spiral (SNAILS): Application to high-resolution diffusion tensor imaging

    Liu C, Bammer R, Kim Dh, and Moseley ME. Self-navigated interleaved spiral (SNAILS): Application to high-resolution diffusion tensor imaging . Magnetic Resonance in Medicine , 52 0 (6): 0 1388--1396, 2004

    work page 2004

  14. [14]

    POCS-enhanced inherent correction of motion-induced phase errors (POCS-ICE) for high-resolution multishot diffusion MRI

    Guo H, Ma X, Zhang Z, Zhang B, Yuan C, and Huang F. POCS-enhanced inherent correction of motion-induced phase errors (POCS-ICE) for high-resolution multishot diffusion MRI . Magnetic Resonance in Medicine , 75 0 (1): 0 169--180, 2016

    work page 2016

  15. [15]

    POCS-based reconstruction of multiplexed sensitivity encoded MRI (POCSMUSE): A general algorithm for reducing motion-related artifacts

    Chu ML, Chang HC, Chung HW, Truong TK, Bashir MR, and Chen NK. POCS-based reconstruction of multiplexed sensitivity encoded MRI (POCSMUSE): A general algorithm for reducing motion-related artifacts. Magnetic Resonance in Medicine , 74 0 (5): 0 1336--48, 2015

    work page 2015

  16. [16]

    O'Halloran RL, Holdsworth S, Aksoy M, and Bammer R. Model for the correction of motion-induced phase errors in multishot diffusion-weighted-MRI of the head: Are cardiac-motion-induced phase errors reproducible from beat-to-beat? Magnetic Resonance in Medicine , 68 0 (2): 0 430--440, 2012

    work page 2012

  17. [17]

    SENSE: sensitivity encoding for fast MRI

    Pruessmann KP, Weiger M, Scheidegger MB, and Boesiger P. SENSE: sensitivity encoding for fast MRI. Magnetic resonance in medicine , 42 0 (5): 0 952--62, 1999

    work page 1999

  18. [18]

    Multiplier methods: A survey

    Bertsekas DP. Multiplier methods: A survey . Automatica , 12 0 (2): 0 133--145, 1976

    work page 1976

  19. [19]

    Low-rank Matrix Recovery via Iteratively Reweighted Least Squares Minimization

    Fornasier M, Rauhut H, and Ward R. Low-rank Matrix Recovery via Iteratively Reweighted Least Squares Minimization . SIAM Journal on Optimization , 21 0 (4): 0 1614--1640, 2011

    work page 2011

  20. [20]

    Parallel MR image reconstruction using augmented Lagrangian methods

    Ramani S and Fessler JA. Parallel MR image reconstruction using augmented Lagrangian methods. IEEE Trans. Med. Imaging. , 30 0 (3): 0 694--706, 2011

    work page 2011

  21. [21]

    A Fast Algorithm for Convolutional Structured Low-Rank Matrix Recovery

    Ongie G and Jacob M. A Fast Algorithm for Convolutional Structured Low-Rank Matrix Recovery . IEEE Transactions on Computational Imaging , 2017

    work page 2017

  22. [22]

    Iteratively reweighted algorithms for compressive sensing

    Chartrand R and Wotao Yin . Iteratively reweighted algorithms for compressive sensing . In 2008 IEEE International Conference on Acoustics, Speech and Signal Processing , pages 3869--3872. IEEE, 2008. ISBN 978-1-4244-1483-3

    work page 2008

  23. [23]

    Iterative Reweighted Least Squares for Matrix Rank Minimization

    Mohan K and Fazel M. Iterative Reweighted Least Squares for Matrix Rank Minimization . In 2010 48th Annual Allerton Conference on Communication, Control, and Computing (Allerton) , 2010. ISBN 9781424482160

    work page 2010

  24. [24]

    Virtual coil concept for improved parallel MRI employing conjugate symmetric signals

    Blaimer M, Gutberlet M, Kellman P, Breuer FA, K \" o stler H, and Griswold MA. Virtual coil concept for improved parallel MRI employing conjugate symmetric signals . Magnetic Resonance in Medicine , 61 0 (1): 0 93--102, 2009

    work page 2009

  25. [25]

    LORAKS makes better SENSE: Phase-constrained partial fourier SENSE reconstruction without phase calibration

    Kim TH, Setsompop K, and Haldar JP. LORAKS makes better SENSE: Phase-constrained partial fourier SENSE reconstruction without phase calibration. Magnetic resonance in medicine , 2016

    work page 2016

  26. [26]

    Advances in diffusion MRI acquisition and processing in the Human Connectome Project

    Sotiropoulos SN, Jbabdi S, Xu J, Andersson JL, Moeller S, Auerbach EJ, Glasser MF, Hernandez M, Sapiro G, Jenkinson M, Feinberg DA, Yacoub E, Lenglet C, Van Essen DC, Ugurbil K, and Behrens TEJ. Advances in diffusion MRI acquisition and processing in the Human Connectome Project . NeuroImage , 80: 0 125--143, 2013

    work page 2013

  27. [27]

    High-resolution in vivo diffusion imaging of the human brain with generalized slice dithered enhanced resolution: Simultaneous multislice (gSlider-SMS)

    Setsompop K, Fan Q, Stockmann J, Bilgic B, Huang S, Cauley SF, Nummenmaa A, Wang F, Rathi Y, Witzel T, and Wald LL. High-resolution in vivo diffusion imaging of the human brain with generalized slice dithered enhanced resolution: Simultaneous multislice (gSlider-SMS) . Magnetic Resonance in Medicine , 79 0 (1): 0 141--151, 2018

    work page 2018

  28. [28]

    High slew-rate head-only gradient for improving distortion in echo planar imaging: Preliminary experience

    Tan ET, Lee SK, Weavers PT, Graziani D, Piel JE, Shu Y, Huston J, Bernstein MA, and Foo TK. High slew-rate head-only gradient for improving distortion in echo planar imaging: Preliminary experience . Journal of Magnetic Resonance Imaging , 44 0 (3): 0 653--664, 2016

    work page 2016

  29. [29]

    k-space and q-space: combining ultra-high spatial and angular resolution in diffusion imaging using ZOOPPA at 7 T

    Heidemann RM, Anwander A, Feiweier T, Kn \" o sche TR, and Turner R. k-space and q-space: combining ultra-high spatial and angular resolution in diffusion imaging using ZOOPPA at 7 T. NeuroImage , 60 0 (2): 0 967--78, 2012

    work page 2012

  30. [30]

    SMS MUSSELS: A Navigator-free Reconstruction for Simultaneous MultiSlice Accelerated MultiShot Diffusion Weighted Imaging

    Mani M, Jacob M, McKinnon G, Yang B, Rutt B, Kerr A, and Magnotta V. SMS MUSSELS: A Navigator-free Reconstruction for Simultaneous MultiSlice Accelerated MultiShot Diffusion Weighted Imaging . 2019

    work page 2019