pith. sign in

arxiv: 2501.09799 · v6 · pith:CBLFW63Wnew · submitted 2025-01-16 · 📡 eess.IV

Scan-Adaptive MRI Undersampling Using Neighbor-based Optimization (SUNO)

Siddhant Gautam , Angqi Li , Nicole Seiberlich , Jeffrey A. Fessler , Saiprasad Ravishankar This is my paper

Pith reviewed 2026-05-23 04:58 UTC · model grok-4.3

classification 📡 eess.IV
keywords MRI reconstructionadaptive undersamplingCartesian k-space samplingnearest-neighbor selectionscan-adaptive patternsfastMRI datasetacceleration factorsiterative coordinate descent
0
0 comments X

The pith

Jointly learned scan-adaptive undersampling patterns selected by nearest-neighbor lookup from low-frequency data improve MRI reconstruction over fixed patterns at 4x and 8x acceleration.

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

The paper establishes a framework that learns Cartesian undersampling patterns tailored to individual scans and pairs them with a matching reconstruction model. Training uses alternating optimization with iterative coordinate descent to refine patterns for each training example. At inference a nearest-neighbor search on the first acquired low-frequency lines picks the appropriate pattern for a new scan. Experiments on fastMRI knee and brain data show gains in both visual quality and quantitative scores compared with standard undersampling masks. A reader would care because shorter scan times with less motion artifact could improve patient comfort and image reliability without changing hardware.

Core claim

By alternating between per-scan optimization of k-space sampling masks via iterative coordinate descent and joint training of a reconstruction network, then selecting the learned mask at test time through nearest-neighbor search on initial low-frequency measurements, the method produces higher-fidelity images than population-level or hand-designed masks at both 4× and 8× acceleration on multi-coil knee and brain data.

What carries the argument

The SUNO alternating optimization that pairs ICD-based per-example sampling-pattern learning with nearest-neighbor selection from low-frequency k-space at test time.

If this is right

  • Higher visual quality and quantitative metrics than standard undersampling at both 4× and 8× acceleration on fastMRI knee and brain data.
  • The same training procedure yields patterns that work for multi-coil acquisitions without additional hardware changes.
  • The learned patterns remain Cartesian and therefore compatible with existing scanner trajectories.
  • Code release allows direct replication on other multi-coil datasets.

Where Pith is reading between the lines

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

  • The approach may reduce reliance on very large population datasets because each scan receives its own pattern rather than a single average mask.
  • The neighbor search could be replaced by a small learned classifier if low-frequency features prove insufficient for some anatomies.
  • Extension to non-Cartesian trajectories would require only replacing the ICD step while keeping the neighbor-selection logic.

Load-bearing premise

A nearest-neighbor lookup that uses only the initially acquired low-frequency k-space lines is sufficient to pick an effective scan-specific sampling pattern for unseen test scans.

What would settle it

A test set where the nearest-neighbor-chosen patterns produce higher reconstruction error or visibly worse images than a single fixed mask on a statistically meaningful fraction of cases.

Figures

Figures reproduced from arXiv: 2501.09799 by Angqi Li, Jeffrey A. Fessler, Nicole Seiberlich, Saiprasad Ravishankar, Siddhant Gautam.

Figure 2
Figure 2. Figure 2: Alternating framework for mask and reconstructor [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 1
Figure 1. Figure 1: Illustration of scan-adaptive undersampling patterns [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Schematic of offline iterative coordinate descent (ICD) [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of different masks used for reconstruction at a) [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Reconstructed images using MoDL reconstruction network at [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Reconstructed images using MoDL reconstruction network at [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Reconstructed and error images using different undersampling patterns with the MoDL reconstruction network (two [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Reconstructed and error images using different undersampling patterns using MoDL reconstruction network(two-channel) [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Comparing reconstructed images using SUNO masks initial￾ized from uniform random and LOUPE masks at 8× acceleration factor. Initial Mask Chosen NRMSE ↓ SSIM ↑ PSNR ↑ Uniform Random 0.164 0.896 28.45 LOUPE 0.142 0.903 29.78 TABLE VII: Mean reconstruction metrics for masks initialized with uniform random and LOUPE at an 8× acceleration factor, evaluated over 50 test cases. Initializing with LOUPE results in … view at source ↗
Figure 10
Figure 10. Figure 10: Comparison of reconstructed images from masks optimized using 1) U-Net and 2) MoDL as the reconstruction model. For each mask, the reconstructed images using both the U-Net and MoDL networks is shown. G. Computational Cost of Proposed Approaches This section discusses the time complexity of our proposed algorithms: 1) the sampling optimization algorithm and 2) the nearest neighbor search. 1) Complexity of… view at source ↗
read the original abstract

Accelerated MRI involves collecting partial $k$-space measurements to reduce acquisition time, patient discomfort, and motion artifacts, and typically uses regular undersampling patterns or human-designed schemes. Recent works have studied population-adaptive sampling patterns learned from a group of patients (or scans). However, such patterns can be sub-optimal for individual scans, as they may fail to capture scan or slice-specific details, and their effectiveness can depend on the size and composition of the population. To overcome this issue, we propose a framework for jointly learning scan-adaptive Cartesian undersampling patterns and a corresponding reconstruction model from a training set. We use an alternating algorithm for learning the sampling patterns and the reconstruction model where we use an iterative coordinate descent (ICD) based offline optimization of scan-adaptive $k$-space sampling patterns for each example in the training set. A nearest neighbor search is then used to select the scan-adaptive sampling pattern at test time from initially acquired low-frequency $k$-space information. We applied the proposed framework (dubbed SUNO) to the fastMRI multi-coil knee and brain datasets, demonstrating improved performance over the currently used undersampling patterns at both $4\times$ and $8\times$ acceleration factors in terms of both visual quality and quantitative metrics. The code for the proposed framework is available at https://github.com/sidgautam95/adaptive-sampling-mri-suno. This paper has been accepted for publication in IEEE Transactions on Computational Imaging. The final published version is available at https://doi.org/10.1109/TCI.2026.3653330.

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

2 major / 2 minor

Summary. The manuscript introduces SUNO, a framework that jointly learns scan-adaptive Cartesian undersampling masks and a reconstruction network. An iterative coordinate descent procedure optimizes one mask per training scan offline; at test time a nearest-neighbor lookup in the initially acquired low-frequency k-space coefficients selects the mask for an unseen scan. Experiments on fastMRI multi-coil knee and brain data report improved visual and quantitative reconstruction quality relative to standard fixed undersampling patterns at both 4× and 8× acceleration.

Significance. If the nearest-neighbor selection reliably retrieves a mask whose reconstruction benefit exceeds that of any single fixed mask, the method would constitute a practical advance over population-level adaptive sampling. The public release of code is a clear strength that supports reproducibility and further investigation.

major comments (2)
  1. [Method (nearest-neighbor selection) and Results (quantitative comparisons)] The central empirical claim—that NN lookup on low-frequency coefficients yields a scan-specific advantage—requires evidence that Euclidean (or other) distance in the low-frequency subspace is correlated with mask-specific reconstruction gain on unseen anatomy. No such correlation analysis, ablation against the single best fixed mask from the training collection, or comparison to random selection appears in the method or results sections; without it the reported gains could be explained by the performance of the single strongest training mask.
  2. [Section 3 (alternating algorithm)] The alternating optimization alternates between mask learning via ICD and network training, yet no convergence analysis or sensitivity study to the number of ICD iterations or the size of the training set is supplied. If the learned masks are sensitive to these choices, the claimed superiority at 4× and 8× may not generalize.
minor comments (2)
  1. [Abstract and Results] The abstract states that performance is demonstrated “in terms of both visual quality and quantitative metrics” but does not name the metrics (SSIM, PSNR, etc.) or report numerical values; the results section should make these explicit in the main text or a table.
  2. [Method] Notation for the low-frequency subspace used in the NN search should be defined once and used consistently; currently the description mixes “initially acquired low-frequency k-space information” with later references to the same quantity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful review and constructive comments. We address each major comment below and outline the revisions we plan to make.

read point-by-point responses

  1. Referee: The central empirical claim—that NN lookup on low-frequency coefficients yields a scan-specific advantage—requires evidence that Euclidean (or other) distance in the low-frequency subspace is correlated with mask-specific reconstruction gain on unseen anatomy. No such correlation analysis, ablation against the single best fixed mask from the training collection, or comparison to random selection appears in the method or results sections; without it the reported gains could be explained by the performance of the single strongest training mask.

    Authors: We acknowledge the importance of validating that the NN selection provides a scan-specific benefit beyond what a single fixed mask could achieve. The manuscript demonstrates improvements over standard fixed patterns (e.g., equispaced), but does not include the suggested ablations. We will add these analyses in the revision: a correlation study between low-frequency distances and reconstruction gains, comparison to the best single mask from the training set, and random selection baselines. This will strengthen the evidence for the adaptive approach. revision: yes

  2. Referee: The alternating optimization alternates between mask learning via ICD and network training, yet no convergence analysis or sensitivity study to the number of ICD iterations or the size of the training set is supplied. If the learned masks are sensitive to these choices, the claimed superiority at 4× and 8× may not generalize.

    Authors: The alternating algorithm in Section 3 is an empirical procedure without a provided convergence guarantee. We will include an empirical sensitivity analysis to the number of ICD iterations and training set size in the revised version to address concerns about generalization. Regarding formal convergence analysis, this may be challenging due to the non-convex nature of the joint optimization, but the empirical results support practical utility. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical optimization pipeline evaluated externally

full rationale

The paper presents an alternating ICD-based optimization to learn per-scan sampling masks from training data, followed by NN lookup on low-frequency k-space at test time. No equations, predictions, or uniqueness claims reduce by construction to fitted inputs or self-citations; the central claim is an empirical performance improvement on held-out fastMRI knee/brain data. The derivation chain is a standard supervised learning pipeline with external validation and does not invoke self-referential definitions or load-bearing prior results from the same authors.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that low-frequency k-space data suffices for pattern selection and on the empirical effectiveness of the alternating ICD optimization; both are unverified in the abstract alone.

axioms (1)
  • domain assumption Low-frequency k-space measurements contain enough information to identify the most suitable scan-specific undersampling pattern via nearest-neighbor lookup.
    This premise underpins the test-time selection step described in the abstract.

pith-pipeline@v0.9.0 · 5841 in / 1219 out tokens · 42643 ms · 2026-05-23T04:58:45.447802+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

  • IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    We use an alternating algorithm for learning the sampling patterns and the reconstruction model where we use an iterative coordinate descent (ICD) based offline optimization of scan-adaptive k-space sampling patterns for each example in the training set. A nearest neighbor search is then used to select the scan-adaptive sampling pattern at test time from initially acquired low-frequency k-space information.

  • IndisputableMonolith/Foundation/AbsoluteFloorClosure.lean reality_from_one_distinction unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    The proposed SUNO framework... demonstrating improved performance over the currently used undersampling patterns at both 4× and 8× acceleration factors

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

71 extracted references · 71 canonical work pages · 4 internal anchors

  1. [1]

    Liang and P.C

    Z. Liang and P.C. Lauterbur,Principles of magnetic resonance Imaging, SPIE Optical Engineering Press Belllingham, W A, 2000

    work page 2000

  2. [2]

    Bernstein, K.F

    M.A. Bernstein, K.F. King, and X.J. Zhou,Handbook of MRI pulse sequences, Elsevier, 2004

    work page 2004

  3. [3]

    Ultrafast imaging: principles, pitfalls, solutions, and appli- cations,

    J. Tsao, “Ultrafast imaging: principles, pitfalls, solutions, and appli- cations,”Journal of Magnetic Resonance Imaging, vol. 32, no. 2, pp. 252–266, 2010

    work page 2010

  4. [4]

    SENSE: sensitivity encoding for fast MRI,

    K.P. Pruessmann, M. Weiger, M.B. Scheidegger, and P. Boesiger, “SENSE: sensitivity encoding for fast MRI,”Magnetic Resonance in Medicine, vol. 42, no. 5, pp. 952–962, 1999

    work page 1999

  5. [5]

    Generalized autocalibrating partially parallel acquisitions (grappa),

    Mark A Griswold, Peter M Jakob, Robin M Heidemann, Mathias Nittka, Vladimir Jellus, Jianmin Wang, Berthold Kiefer, and Axel Haase, “Generalized autocalibrating partially parallel acquisitions (grappa),” Magnetic Resonance in Medicine: An Official Journal of the Interna- tional Society for Magnetic Resonance in Medicine, vol. 47, no. 6, pp. 1202–1210, 2002

    work page 2002

  6. [6]

    Parallel MRI using phased array coils,

    L. Ying and Z. Liang, “Parallel MRI using phased array coils,”IEEE Signal Processing Magazine, vol. 27, no. 4, pp. 90–98, 2010

    work page 2010

  7. [7]

    Parallel MR imaging,

    A. Deshmane, V . Gulani, M.A. Griswold, and N. Seiberlich, “Parallel MR imaging,”Journal of Magnetic Resonance Imaging, vol. 36, no. 1, pp. 55–72, 2012

    work page 2012

  8. [8]

    Compressed sensing,

    D.L. Donoho, “Compressed sensing,”IEEE Transactions on Information Theory, vol. 52, no. 4, pp. 1289–1306, 2006

    work page 2006

  9. [9]

    Sparse MRI: The application of compressed sensing for rapid MR imaging,

    M. Lustig, D. Donoho, and J.M. Pauly, “Sparse MRI: The application of compressed sensing for rapid MR imaging,”Magnetic Resonance in Medicine, vol. 58, no. 6, pp. 1182–1195, 2007

    work page 2007

  10. [10]

    Compressed sensing MRI,

    M. Lustig, D.L. Donoho, J.M. Santos, and J.M. Pauly, “Compressed sensing MRI,”IEEE signal processing magazine, vol. 25, no. 2, pp. 72–82, 2008

    work page 2008

  11. [11]

    MR image reconstruction from highly undersampled k-space data by dictionary learning,

    S. Ravishankar and Y . Bresler, “MR image reconstruction from highly undersampled k-space data by dictionary learning,”IEEE Transactions on Medicine Imaging, vol. 30, no. 5, pp. 1028–1041, 2010

    work page 2010

  12. [12]

    Blind compressive sensing dynamic MRI,

    S.G. Lingala and M. Jacob, “Blind compressive sensing dynamic MRI,” IEEE Transactions on Medicine Imaging, vol. 32, no. 6, pp. 1132–1145, 2013

    work page 2013

  13. [13]

    Magnetic resonance image reconstruction from undersampled measurements using a patch-based nonlocal operator,

    X. Qu, Y . Hou, F. Lam, D. Guo, J. Zhong, and Z. Chen, “Magnetic resonance image reconstruction from undersampled measurements using a patch-based nonlocal operator,”Medicine image analysis, vol. 18, no. 6, pp. 843–856, 2014

    work page 2014

  14. [14]

    Fast multiclass dictionaries learning with geometrical directions in MRI reconstruction,

    Z. Zhan, J. Cai, D. Guo, Y . Liu, Z. Chen, and X. Qu, “Fast multiclass dictionaries learning with geometrical directions in MRI reconstruction,” IEEE Transactions on biomedical engineering, vol. 63, no. 9, pp. 1850– 1861, 2015

    work page 2015

  15. [15]

    Learning sparsifying transforms,

    S. Ravishankar and Y . Bresler, “Learning sparsifying transforms,”IEEE Transactions on Signal Processing, vol. 61, no. 5, pp. 1072–1086, 2012

    work page 2012

  16. [16]

    Image reconstruction: From sparsity to data-adaptive methods and machine learning,

    S. Ravishankar, J.C. Ye, and J.A. Fessler, “Image reconstruction: From sparsity to data-adaptive methods and machine learning,”Proceedings of the IEEE, vol. 108, no. 1, pp. 86–109, 2019

    work page 2019

  17. [17]

    Deep ADMM-Net for compressive sensing MRI,

    J. Sun, H. Li, Z. Xu, et al., “Deep ADMM-Net for compressive sensing MRI,”Advances in neural information processing systems, vol. 29, 2016

    work page 2016

  18. [18]

    ADMM-Net: A Deep Learning Approach for Compressive Sensing MRI

    Y . Yang, J. Sun, H. Li, and Z. Xu, “ADMM-Net: A deep learning ap- proach for compressive sensing MRI,”arXiv preprint arXiv:1705.06869, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  19. [19]

    Ista-net: Interpretable optimization- inspired deep network for image compressive sensing,

    Jian Zhang and Bernard Ghanem, “Ista-net: Interpretable optimization- inspired deep network for image compressive sensing,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 1828–1837

    work page 2018

  20. [20]

    U-Net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional networks for biomedical image segmentation,” inInternational Confer- ence on Medicine image computing and computer-assisted intervention. Springer, 2015, pp. 234–241

    work page 2015

  21. [21]

    Deep learning for undersampled MRI reconstruction,

    C.M. Hyun, H.P. Kim, S.M. Lee, S. Lee, and J.K. Seo, “Deep learning for undersampled MRI reconstruction,”Physics in Medicine & Biology, vol. 63, no. 13, pp. 135007, 2018

    work page 2018

  22. [22]

    Image reconstruction by domain-transform manifold learning,

    B. Zhu, J.Z. Liu, S.F. Cauley, B.R. Rosen, and M.S. Rosen, “Image reconstruction by domain-transform manifold learning,”Nature, vol. 555, no. 7697, pp. 487–492, 2018

    work page 2018

  23. [23]

    Image reconstruction is a new frontier of machine learning,

    G. Wang, J.C. Ye, K. Mueller, and J.A. Fessler, “Image reconstruction is a new frontier of machine learning,”IEEE Transactions on Medicine Imaging, vol. 37, no. 6, pp. 1289–1296, 2018

    work page 2018

  24. [24]

    Deep-learning methods for parallel magnetic resonance imaging reconstruction: A survey of the current approaches, trends, and issues,

    Florian Knoll, Kerstin Hammernik, Chi Zhang, Steen Moeller, Thomas Pock, Daniel K Sodickson, and Mehmet Akcakaya, “Deep-learning methods for parallel magnetic resonance imaging reconstruction: A survey of the current approaches, trends, and issues,”IEEE signal processing magazine, vol. 37, no. 1, pp. 128–140, 2020

    work page 2020

  25. [25]

    Learning a variational network for reconstruction of accelerated MRI data,

    K. Hammernik, T. Klatzer, E. Kobler, M.P. Recht, D.K. Sodickson, T. Pock, and F. Knoll, “Learning a variational network for reconstruction of accelerated MRI data,”Magnetic Resonance in Medicine, vol. 79, no. 6, pp. 3055–3071, 2018

    work page 2018

  26. [26]

    End-to-end variational networks for acceler- ated MRI reconstruction,

    A. Sriram, J. Zbontar, T. Murrell, A. Defazio, C.L. Zitnick, N. Yakubova, F. Knoll, and P. Johnson, “End-to-end variational networks for acceler- ated MRI reconstruction,” inMedicine Image Computing and Computer Assisted Intervention–MICCAI 2020: 23rd International Conference, Lima, Peru, October 4–8, 2020, Proceedings, Part II 23. Springer, 2020, pp. 64–73

    work page 2020

  27. [27]

    Deep Generative Adversarial Networks for Compressed Sensing Automates MRI

    M. Mardani, E. Gong, J.Y . Cheng, S. Vasanawala, G. Zaharchuk, M. Al- ley, N. Thakur, S. Han, W. Dally, J.M. Pauly, et al., “Deep generative adversarial networks for compressed sensing automates MRI,”arXiv preprint arXiv:1706.00051, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  28. [28]

    Dagan: deep de-aliasing generative adversarial networks for fast compressed sensing MRI reconstruction,

    G. Yang, S. Yu, H. Dong, G. Slabaugh, P.L. Dragotti, X. Ye, F. Liu, S. Arridge, J. Keegan, Y . Guo, et al., “Dagan: deep de-aliasing generative adversarial networks for fast compressed sensing MRI reconstruction,” IEEE Transactions on Medicine Imaging, vol. 37, no. 6, pp. 1310–1321, 2017

    work page 2017

  29. [29]

    MoDL: Model-based deep learning architecture for inverse problems,

    H.K. Aggarwal, M.P. Mani, and M. Jacob, “MoDL: Model-based deep learning architecture for inverse problems,”IEEE Transactions on Medicine Imaging, vol. 38, no. 2, pp. 394–405, 2018

    work page 2018

  30. [30]

    J-MoDL: Joint model-based deep learning for optimized sampling and reconstruction,

    H.K. Aggarwal and M. Jacob, “J-MoDL: Joint model-based deep learning for optimized sampling and reconstruction,”IEEE journal of selected topics in signal processing, vol. 14, no. 6, pp. 1151–1162, 2020

    work page 2020

  31. [31]

    Compressed sensing in dynamic MRI,

    U. Gamper, P. Boesiger, and S. Kozerke, “Compressed sensing in dynamic MRI,”Magnetic Resonance in Medicine, vol. 59, no. 2, pp. 365–373, 2008

    work page 2008

  32. [32]

    Compressed-sensing MRI with random encoding,

    J.P. Haldar, D. Hernando, and Z. Liang, “Compressed-sensing MRI with random encoding,”IEEE Transactions on Medicine Imaging, vol. 30, no. 4, pp. 893–903, 2010

    work page 2010

  33. [33]

    Fast poisson disk sampling in arbitrary dimensions.,

    Robert Bridson, “Fast poisson disk sampling in arbitrary dimensions.,” SIGGRAPH sketches, vol. 10, no. 1, pp. 1, 2007. 13

    work page 2007

  34. [34]

    Spirit: iterative self-consistent parallel imaging reconstruction from arbitrary k-space,

    Michael Lustig and John M Pauly, “Spirit: iterative self-consistent parallel imaging reconstruction from arbitrary k-space,”Magnetic resonance in medicine, vol. 64, no. 2, pp. 457–471, 2010

    work page 2010

  35. [35]

    3D cartesian MRI with compressed sensing and variable view sharing using complementary poisson-disc sampling,

    E. Levine, B. Daniel, S. Vasanawala, B. Hargreaves, and M. Saranathan, “3D cartesian MRI with compressed sensing and variable view sharing using complementary poisson-disc sampling,”Magnetic Resonance in Medicine, vol. 77, no. 5, pp. 1774–1785, 2017

    work page 2017

  36. [36]

    Adaptive sampling design for com- pressed sensing MRI,

    S. Ravishankar and Y . Bresler, “Adaptive sampling design for com- pressed sensing MRI,” in2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE, 2011, pp. 3751–3755

    work page 2011

  37. [37]

    Adapted random sampling patterns for accelerated MRI,

    F. Knoll, C. Clason, C. Diwoky, and R. Stollberger, “Adapted random sampling patterns for accelerated MRI,”Magnetic resonance materials in physics, biology and medicine, vol. 24, pp. 43–50, 2011

    work page 2011

  38. [38]

    A robust adaptive sampling method for faster acquisition of MR images,

    J. Vellagoundar and R.R. Machireddy, “A robust adaptive sampling method for faster acquisition of MR images,”Magnetic resonance Imaging, vol. 33, no. 5, pp. 635–643, 2015

    work page 2015

  39. [39]

    Energy preserved sam- pling for compressed sensing MRI,

    Y . Zhang, B.S. Peterson, G. Ji, and Z. Dong, “Energy preserved sam- pling for compressed sensing MRI,”Computational and mathematical methods in medicine, vol. 2014, no. 1, pp. 546814, 2014

    work page 2014

  40. [40]

    OEDIPUS: An experiment design framework for sparsity-constrained MRI,

    J.P. Haldar and D. Kim, “OEDIPUS: An experiment design framework for sparsity-constrained MRI,”IEEE Transactions on Medicine Imaging, vol. 38, no. 7, pp. 1545–1558, 2019

    work page 2019

  41. [41]

    Op- timization of k-space trajectories for compressed sensing by bayesian experimental design,

    M. Seeger, H. Nickisch, R. Pohmann, and Bernhard Sch ¨olkopf, “Op- timization of k-space trajectories for compressed sensing by bayesian experimental design,”Magnetic Resonance in Medicine, vol. 63, no. 1, pp. 116–126, 2010

    work page 2010

  42. [42]

    Learning-based compressive MRI,

    B. G ¨ozc¨u, R. K. Mahabadi, Y . Li, E. Ilıcak, T. Cukur, J. Scarlett, and V . Cevher, “Learning-based compressive MRI,”IEEE Transactions on Medicine Imaging, vol. 37, no. 6, pp. 1394–1406, 2018

    work page 2018

  43. [43]

    Rethinking sampling in parallel MRI: A data-driven approach,

    B. G ¨ozc¨u, T. Sanchez, and V . Cevher, “Rethinking sampling in parallel MRI: A data-driven approach,” in2019 27th European Signal Processing Conference (EUSIPCO). IEEE, 2019, pp. 1–5

    work page 2019

  44. [44]

    Scalable learning-based sampling optimization for compressive dynamic MRI,

    T. Sanchez, B. G ¨ozc¨u, R. van Heeswijk, A. Eftekhari, E. Ilıcak, T. C ¸ ukur, and V . Cevher, “Scalable learning-based sampling optimization for compressive dynamic MRI,” in2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2020, pp. 8584–8588

    work page 2020

  45. [45]

    Fast data-driven learning of parallel MRI sampling patterns for large scale problems,

    M.V . Zibetti, G.T. Herman, and R.R. Regatte, “Fast data-driven learning of parallel MRI sampling patterns for large scale problems,”Scientific Reports, vol. 11, no. 1, pp. 19312, 2021

    work page 2021

  46. [46]

    Alternating learning approach for variational networks and undersampling pattern in parallel MRI applications,

    M.V .W. Zibetti, F. Knoll, and R.R. Regatte, “Alternating learning approach for variational networks and undersampling pattern in parallel MRI applications,”IEEE Transactions on Computational Imaging, vol. 8, pp. 449–461, 2022

    work page 2022

  47. [47]

    Deep- learning-based optimization of the under-sampling pattern in MRI,

    C.D. Bahadir, A.Q. Wang, A.V . Dalca, and M.R. Sabuncu, “Deep- learning-based optimization of the under-sampling pattern in MRI,” IEEE Transactions on Computational Imaging, vol. 6, pp. 1139–1152, 2020

    work page 2020

  48. [48]

    Extending loupe for k-space under-sampling pattern optimization in multi-coil mri,

    J. Zhang, H. Zhang, A. Wang, and et al., “Extending loupe for k-space under-sampling pattern optimization in multi-coil mri,” inMLMIR 2020: Machine Learning for Medical Image Reconstruction. Springer, 2020, pp. 91–101

    work page 2020

  49. [49]

    Learning the sampling pattern for MRI,

    F. Sherry, M. Benning, J.C. De los Reyes, M.J. Graves, G. Maier- hofer, G. Williams, Carola-Bibiane Sch ¨onlieb, and Matthias J Ehrhardt, “Learning the sampling pattern for MRI,”IEEE Transactions on Medicine Imaging, vol. 39, no. 12, pp. 4310–4321, 2020

    work page 2020

  50. [50]

    End- to-end sequential sampling and reconstruction for MR imaging,

    T. Yin, Z. Wu, H. Sun, A.V . Dalca, Y . Yue, and K.L. Bouman, “End- to-end sequential sampling and reconstruction for MR imaging,” in Proceedings of the Machine Learning for Health Conference, 2021

    work page 2021

  51. [51]

    Single-pass object-adaptive data undersampling and reconstruction for MRI,

    Z. Huang and S. Ravishankar, “Single-pass object-adaptive data undersampling and reconstruction for MRI,”IEEE Transactions on Computational Imaging, vol. 8, pp. 333–345, 2022

    work page 2022

  52. [52]

    Autosamp: Autoencoding k-space sampling via variational information maximization for 3D MRI,

    C. Alkan, M. Mardani, C. Liao, Z. Li, S.S. Vasanawala, and J.M. Pauly, “Autosamp: Autoencoding k-space sampling via variational information maximization for 3D MRI,”IEEE Transactions on Medicine Imaging, 2024

    work page 2024

  53. [53]

    Pilot: Physics-informed learned optimized trajectories for accelerated mri,

    T. Weiss, O. Senouf, S. Vedula, O. Michailovich, M. Zibulevsky, and A. Bronstein, “Pilot: Physics-informed learned optimized trajectories for accelerated mri,”arXiv preprint arXiv:1909.05773, 2019

    work page arXiv 1909

  54. [54]

    Sparkling: variable-density k-space filling curves for accelerated T2*-weighted MRI,

    C. Lazarus, P. Weiss, N. Chauffert, F. Mauconduit, L. El Gueddari, C. Destrieux, I. Zemmoura, A. Vignaud, and P. Ciuciu, “Sparkling: variable-density k-space filling curves for accelerated T2*-weighted MRI,”Magnetic Resonance in Medicine, vol. 81, no. 6, pp. 3643–3661, 2019

    work page 2019

  55. [55]

    Optimizing full 3D sparkling trajectories for high-resolution magnetic resonance imaging,

    GR Chaithya, P. Weiss, G. Daval-Fr ´erot, A. Massire, A. Vignaud, and P. Ciuciu, “Optimizing full 3D sparkling trajectories for high-resolution magnetic resonance imaging,”IEEE Transactions on Medical Imaging, vol. 41, no. 8, pp. 2105–2117, 2022

    work page 2022

  56. [56]

    B- spline parameterized joint optimization of reconstruction and k-space trajectories (BJORK) for accelerated 2D MRI,

    G. Wang, T. Luo, J. Nielsen, D.C. Noll, and J.A. Fessler, “B- spline parameterized joint optimization of reconstruction and k-space trajectories (BJORK) for accelerated 2D MRI,”IEEE Transactions on Medicine Imaging, vol. 41, no. 9, pp. 2318–2330, 2022

    work page 2022

  57. [57]

    NC-PDNet: A density-compensated unrolled network for 2D and 3D non-cartesian MRI reconstruction,

    Z. Ramzi, GR Chaithya, J. Starck, and P. Ciuciu, “NC-PDNet: A density-compensated unrolled network for 2D and 3D non-cartesian MRI reconstruction,”IEEE Transactions on Medical Imaging, vol. 41, no. 7, pp. 1625–1638, 2022

    work page 2022

  58. [58]

    Stochastic optimization of three-dimensional non-cartesian sampling trajectory,

    G. Wang, J. F. Nielsen, J. A. Fessler, and D. C. Noll, “Stochastic optimization of three-dimensional non-cartesian sampling trajectory,” Magnetic Resonance in Medicine, vol. 90, no. 2, pp. 417–431, 2023

    work page 2023

  59. [59]

    Active MR k-space sampling with reinforcement learning,

    L. Pineda, S. Basu, A. Romero, R. Calandra, and M. Drozdzal, “Active MR k-space sampling with reinforcement learning,” inMedicine Image Computing and Computer Assisted Intervention–MICCAI 2020: 23rd International Conference, Lima, Peru, October 4–8, 2020, Proceedings, Part II 23. Springer, 2020, pp. 23–33

    work page 2020

  60. [60]

    Experimental design for MRI by greedy policy search,

    T. Bakker, H. van Hoof, and M. Welling, “Experimental design for MRI by greedy policy search,”Advances in Neural Information Processing Systems, vol. 33, pp. 18954–18966, 2020

    work page 2020

  61. [61]

    Image quality assessment: from error visibility to structural similarity,

    Z. Wang, A.C. Bovik, H.R. Sheikh, and E.P. Simoncelli, “Image quality assessment: from error visibility to structural similarity,”IEEE Transactions on image processing, vol. 13, no. 4, pp. 600–612, 2004

    work page 2004

  62. [62]

    Patient-adaptive and learned MRI data undersampling using neighborhood clustering,

    S. Gautam, A. Li, and S. Ravishankar, “Patient-adaptive and learned MRI data undersampling using neighborhood clustering,” in2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2024, pp. 2081–2085

    work page 2024

  63. [63]

    Zero-shot self- supervised learning for MRI reconstruction,

    Burhaneddin Y ., S. A. Hosseini, and M. Akcakaya, “Zero-shot self- supervised learning for MRI reconstruction,” inInternational Confer- ence on Learning Representations, 2022

    work page 2022

  64. [64]

    fastMRI: An open dataset and benchmarks for accelerated MRI,

    J. Zbontar, F. Knoll, A. Sriram, T. Murrell, Z. Huang, M.J. Muckley, A. Defazio, R. Stern, P. Johnson, M. Bruno, et al., “fastMRI: An open dataset and benchmarks for accelerated MRI,”arXiv preprint arXiv:1811.08839, 2018

    work page arXiv 2018

  65. [65]

    fastMRI: A publicly available raw k-space and dicom dataset of knee images for accelerated MR image reconstruction using machine learning,

    F. Knoll, J. Zbontar, A. Sriram, M.J. Muckley, M. Bruno, A. Defazio, M. Parente, K.J. Geras, J. Katsnelson, H. Chandarana, et al., “fastMRI: A publicly available raw k-space and dicom dataset of knee images for accelerated MR image reconstruction using machine learning,” Radiology: Artificial Intelligence, vol. 2, no. 1, pp. e190007, 2020

    work page 2020

  66. [66]

    ESPIRiT — an eigenvalue approach to autocalibrating parallel MRI: where SENSE meets GRAPPA,

    M. Uecker, P. Lai, M.J. Murphy, P. Virtue, M. Elad, J.M. Pauly, S.S. Vasanawala, and M. Lustig, “ESPIRiT — an eigenvalue approach to autocalibrating parallel MRI: where SENSE meets GRAPPA,”Magnetic Resonance in Medicine, vol. 71, no. 3, pp. 990–1001, 2014

    work page 2014

  67. [67]

    Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation

    Yoshua Bengio, Nicholas L ´eonard, and Aaron Courville, “Estimating or propagating gradients through stochastic neurons for conditional computation,”arXiv preprint arXiv:1308.3432, 2013

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  68. [68]

    Deep iterative down-up CNN for image denoising,

    S. Yu, B. Park, and J. Jeong, “Deep iterative down-up CNN for image denoising,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops, 2019

    work page 2019

  69. [69]

    Adam: A Method for Stochastic Optimization

    D.P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  70. [70]

    Adaptive local neighborhood- based neural networks for MR image reconstruction from undersampled data,

    S. Liang, A. Lahiri, and S. Ravishankar, “Adaptive local neighborhood- based neural networks for MR image reconstruction from undersampled data,”IEEE Transactions on Computational Imaging, 2024

    work page 2024

  71. [71]

    fastMRI+, clinical pathol- ogy annotations for knee and brain fully sampled magnetic resonance imaging data,

    R. Zhao, B. Yaman, Y . Zhang, R. Stewart, A. Dixon, F. Knoll, Z. Huang, Y . W. Lui, M. S. Hansen, and M. P. Lungren, “fastMRI+, clinical pathol- ogy annotations for knee and brain fully sampled magnetic resonance imaging data,”Scientific Data, vol. 9, no. 1, pp. 152, 2022

    work page 2022