Pith. sign in

REVIEW 3 major objections 3 minor 66 references

DUN-SRE: Deep Unrolling Network with Spatiotemporal Rotation Equivariance for Dynamic MRI Reconstruction

T0 review · 3 major / 3 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A deep unrolling network that bakes rotation symmetry into both space and time improves dynamic MRI reconstruction under heavy undersampling.

desk verdict Strong empirical results on cardiac cine MRI, but the 'rigorous spatiotemporal rotation equivariance' claim outruns the actual construction—worth refereeing, with revision demands. read the letter →

arxiv 2506.10309 v1 pith:TZ367Z7C submitted 2025-06-12 eess.IV cs.AIcs.CV

Yuliang Zhu , Jing Cheng , Qi Xie , Zhuo-Xu Cui , Qingyong Zhu , Yuanyuan Liu , Xin Liu , Jianfeng Ren
show 2 more authors
Chengbo Wang Dong Liang
This is my paper · ORCID
classification eess.IVcs.AIcs.CV
keywords dynamicMRIreconstructionrotationequivariancespatiotemporalsymmetrydeepunrollingnetworkcardiaccineequivariantconvolutiondataconsistencyfilterparameterization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that dynamic MRI reconstruction improves when the network architecture itself enforces rotation symmetry in both space and time, not just within individual frames. It proposes DUN-SRE, a deep unrolling network whose data-consistency and proximal modules are both built from spatiotemporal rotation-equivariant convolutions, so the symmetry constraint propagates through every iteration. On cardiac cine datasets it reports consistent gains over prior methods at all tested acceleration factors, with the largest improvements at high undersampling and in rotation-symmetric structures such as myocardial boundaries and papillary muscles. If the claim holds, this offers a parameter-efficient way to make dynamic reconstruction robust to aggressive undersampling without relying on data augmentation alone.

What carries the argument

The load-bearing object is the SREC (Spatiotemporal Rotation Equivariant Convolution) module, a (2+1)D equivariant convolution that separates spatial and temporal processing. Spatial layers apply group-equivariant convolutions over the rotation group G (set to four pi/2 rotations), producing feature maps indexed by rotation; the temporal layer is a pseudo-1D convolution with 1x1 spatial support acting on a group SG that is effectively trivial but still performs cyclic shifts of the group sub-channels, which transfers equivariance between consecutive 2D layers. The same SREC block builds both the proximal mapping network and a learned data-consistency network, so rotation equivariance is preserved through every unrolled iteration. A Fourier-series filter parameterization replaces naive filter rotation to avoid interpolation artifacts, and the paper argues this combined design is what prior unrolling networks lacked.

What would settle it

Train DUN-SRE but replace the temporal equivariant layer's cyclic channel shift with an ordinary 1D convolution of the same receptive field and parameter count; if PSNR, SSIM, and HFEN on the in-house cine dataset at R=20 and R=24 do not degrade, the reported gains do not come from temporal rotation equivariance.

Watch

Extended reading notes

Core claim

The central claim is that rotation equivariance must be modeled jointly in the spatial and temporal dimensions for dynamic MRI, and that this can be done inside an unrolled optimization network. Prior equivariant CNNs cover planar or volumetric rotations, but a naive temporal convolution breaks the group structure established by the 2D layers. DUN-SRE introduces a (2+1)D spatiotemporal rotation equivariant convolution (SREC) that factorizes the operation into 2D group-equivariant spatial layers and a dedicated 1D temporal equivariant layer, with cyclic group-channel shifts maintaining equivariance across time. Because both the proximal network and the data-consistency network use SREC, the entire unrolled network is equivariant end to end, and a Fourier-basis filter parameterization keeps the rotated kernel representations accurate. The paper reports quantitative gains over five prior methods on its in-house cine dataset across R=8, 12, 16, 20, and 24, and visual improvements in the temporal (y-t) domain as well as on the OCMR generalization and prospective data.

Load-bearing premise

The construction assumes that cardiac dynamics can be faithfully modeled by a rotation-equivariant structure in which the temporal layer only cyclically shifts rotation-labeled channels, rather than requiring a more general motion model.

Editorial extensions

If this is right

  • Rotation symmetry in dynamic reconstruction is best enforced by architecture rather than by data augmentation, since the unrolled network remains equivariant at every iteration.
  • The equivariant data-consistency module is load-bearing: ablations show that replacing it with a conventional CNN-based DC term degrades performance and breaks end-to-end equivariance.
  • Weight sharing across rotation group and time means the symmetry prior improves parameter efficiency, which should help in low-data regimes.
  • The method generalizes across scanners and to prospectively undersampled data without fine-tuning, suggesting the symmetry prior transfers across acquisition domains.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The practical content of the temporal equivariance is a cyclic shift of rotation-labeled channels across frames, so the physical interpretation as 'rotation over time' is looser than the name suggests; the benefit may come from enforcing group coherence and weight sharing rather than from a true temporal rotation model.
  • The same (2+1)D equivariant construction should transfer to other dynamic inverse problems, such as video super-resolution or motion-resolved imaging, whenever frames contain orientation-varied repeating structures.
  • Because the equivariance is discrete (pi/2 rotations), part of the gain may come from matching the four-fold symmetry of Cartesian sampling and the roughly circular heart cross-section; testing N_G=8 or continuous steerable filters would reveal how much finer rotational symmetry adds.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 3 minor

Summary. The paper proposes DUN-SRE, a deep unrolling network for dynamic MRI reconstruction that aims to embed rotation equivariance in both the spatial and temporal dimensions via a (2+1)D equivariant convolution architecture. The network consists of a spatiotemporal rotation equivariant proximal module (SREC-Prox) and an equivariant data consistency module (SREC-DC), along with a Fourier-basis filter parametrization intended to preserve representation fidelity. The method is evaluated on an in-house cardiac cine dataset and the public OCMR dataset, reporting state-of-the-art PSNR, SSIM, and HFEN at acceleration factors R=8,12,16,20,24, with ablation studies on the proximal module, the DC module, and the filter parametrization.

Significance. If the equivariance claims were correct, DUN-SRE would be a notable contribution to symmetry-aware deep learning for dynamic imaging, potentially improving reconstruction quality at high acceleration by leveraging a physically motivated prior. The paper provides a substantial experimental comparison with multiple baselines, per-metric means and standard deviations, and a 3-fold cross-validation ablation, which are strengths. However, the central theoretical claim of rigorous spatiotemporal rotation equivariance is undermined by the trivial temporal group and the non-commuting forward operator in the data consistency term; the empirical gains may still stand, but they cannot be attributed to exact spatiotemporal equivariance as the paper states.

major comments (3)
  1. [III-C2 (Remark 1)] Remark 1 claims that an L2-norm data consistency update is equivariant to any unitary transformation ρ. This assertion is only valid if the forward operator A commutes with ρ and if the measured data y transforms accordingly under the group action. For the actual acquisition A = PΩFS with the Cartesian VISTA mask, the sampling mask PΩ is anisotropic and does not commute with 90° rotations: a mask that samples only vertical k-space lines, applied to a rotated image, still samples the same vertical lines rather than the rotated set. Consequently the gradient descent step x − η A^H(Ax − y) is not rotation-equivariant. The SREC-DC network, which takes the raw k-space data and ESPIRiT sensitivity maps as inputs, inherits this non-equivariance unless it entirely ignores y and the mask, which would sever data consistency. This invalidates the global equivariance assertion in Eq. (15).
  2. [III-B4] The temporal equivariant layer is defined with a group SG that is explicitly stated to act as a trivial group in practical implementation, composed only of the identity element (Section III-B4). A trivial group action does not implement any temporal rotation symmetry; the cyclic channel shifts only permute the spatial group index across time and do not rotate the temporal axis. Therefore the network enforces spatial rotation equivariance only, and the claims of 'spatiotemporal rotation equivariance' and 'temporal rotation symmetry constraints' in the abstract and introduction are not supported. The empirical gains attributed to temporal equivariance could instead arise from the (2+1)D factorization and weight sharing across frames. The authors should either implement a non-trivial temporal group action or revise the framing to channel-level group shuffling and temper the symmetry claims accordingly.
  3. [III-C2 (Eq. (15))] The derivation of global equivariance is only a chain of equalities with no formal proof or numerical verification. Equation (15) is also not well-typed: applying Q_SREC to ρ_g and then to X^k is ambiguous, and the composition of operators is not clearly defined. More fundamentally, even if H_SREC and Q_SREC were each equivariant, the composition is equivariant only if the group action on every input to the DC network, including the measured k-space data y and the ESPIRiT sensitivity maps, is compatible with the image-space rotation. The manuscript does not define such an action, and for fixed y and maps the composition cannot be equivariant. We suggest explicitly stating the group action on all inputs and providing an equivariance error measurement (e.g., applying random 90° rotations to test inputs and measuring the output disagreement) to support the claim.
minor comments (3)
  1. [Table III] At R=24, DL-ESPIRiT and U-MCMR report identical PSNR (38.6397) and SSIM (0.9373) values but different HFEN values; this appears to be a transcription or copying error and should be corrected.
  2. [III-B4] There are several typos in the manuscript, including 'imtermediate' in Section III-B4, 'dicrete' in Section III-B1, 'denonted' in Section III-C1, and 'nework' in the Conclusion; these should be fixed.
  3. [III-D] Equation (17) appears to have a missing closing parenthesis in the definition of the 1D Fourier basis filter, and the summation index and limits should be checked for consistency with the filter size p.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: DUN-SRE's central results are empirical architecture comparisons, and its self-citations are background or baseline references rather than load-bearing inputs.

full rationale

Walking the derivation chain, the paper builds an unrolled proximal-gradient network (Eq. 4), implements the proximal and data-consistency modules with the proposed SREC layers (Sec. III-C1 and III-C2), and then reports performance gains from direct comparisons on held-out cine data (Tab. III, Fig. 5-10). The equivariant convolution construction (Eqs. 5-12) follows standard group-convolution formalism and external ECNN literature [25, 28, 30, 31], not a self-referential uniqueness theorem. The self-citations [35, 52, 65] are a preliminary-study reference, a learned-DC motivation reference, and a comparison baseline, respectively; none is used as an unverified premise that forces the main conclusion. The filter parameterization in Sec. III-D is an architectural design choice, not a fitted parameter that is later renamed as a prediction. The strongest concern in the paper is non-circular: Remark 1 (Sec. III-C2) asserts L2-DC equivariance under a unitary transformation without stating the required hypothesis that the forward operator A commutes with the rotation; for the Cartesian VISTA mask used in the experiments, the mask does not commute with 90-degree rotations, so the 'rigorous propagation' claim is a correctness and soundness gap rather than a circular reduction. Similarly, Sec. III-B4 states that SG acts as a trivial group, making the temporal rotation symmetry claim nominal; this is a terminological or architectural mismatch, not a derivation that reduces to its own input. No prediction is equivalent by construction to a fitted constant or to a self-citation chain, so the circularity burden is low.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The central claim rests on a small number of explicit design choices: a discrete rotation group G of size 4, an unrolled depth of 10, a 1/4 channel reduction, and a Fourier basis filter representation. These are hyperparameters rather than derived constants. The main domain assumption, that cardiac cine frames contain repeated rotated anatomy across space and time, is plausible but not quantitatively verified. The trickiest load-bearing premise is that the temporal equivariant layer, whose group SG is the identity in implementation, preserves the equivariance of the 2D layers; this is asserted rather than proven. The architecture does not introduce new physical entities; SG is a notational device.

free parameters (4)
  • Rotation group size N_G = 4 (pi/2 rotations)
    Set in Sec. IV-A2; defines the discrete symmetry group and channel expansion used by SREC.
  • Unrolling iterations K = 10
    Set empirically in Sec. IV-A2; depth of the equivariance propagation in DUN-SRE.
  • Channel reduction factor for equivariant layers = 1/4 of original channel count
    Set in Sec. IV-A2 to keep memory use comparable to baselines; affects representational capacity.
  • Fourier basis count for filter parameterization = Not reported
    Sec. III-D defines filters as a linear combination of basis functions but does not state the number of basis functions used in experiments.
assumptions (5)
  • standard math Image and feature maps are modeled as smooth functions on R^2 x R and SE(2) x R, with Haar measure group convolutions on SO(2).
    Used in the continuous-domain definitions in Eqs. (5), (7), (9) and Sec. III-B, following Cohen and Welling [25].
  • domain assumption Dynamic cardiac cine MRI contains planar rotation symmetry within frames and repeated rotated anatomy across time frames.
    This prior motivates the SREC design in Sec. I and Fig. 1; if real cine data do not respect it, the equivariance constraint is a bias rather than a benefit.
  • ad hoc to paper The learned SREC-DC network preserves the equivariance of the unrolled iterations, not just the explicit L2 DC term.
    Remark 1 proves equivariance only for the L2-norm DC term; Eq. (15) asserts the same for the learned SREC-DC network without a proof.
  • standard math Pointwise nonlinearities such as LeakyReLU and the linear layers are equivariant under the discrete rotation group G, and discretization does not break this.
    Invoked in Sec. III-C1 following [25], [31]; true for continuous group actions but approximate under finite grids and interpolated rotations.
  • domain assumption ESPIRiT sensitivity maps computed from time-averaged k-space data provide a sufficiently accurate forward model for training and testing.
    The pipeline in Sec. IV-A2 and Fig. 4 uses ESPIRiT maps without sensitivity-map error modeling.
invented entities (1)
  • SG, the special group for temporal equivariance
    purpose: Transfer equivariance between 2D equivariant layers through cyclic shifts of group channels in the 1D temporal convolution.
    Defined in Sec. III-B4; in implementation SG 'acts as a trivial group' and is a single identity element, so it is a mathematical bookkeeping device without independent empirical support.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DUN-SRE: Deep Unrolling Network with Spatiotemporal Rotation Equivariance for Dynamic MRI Reconstruction." pith.science (2026). https://pith.science/paper/TZ367Z7C

@misc{pith2026250610309,
  author       = {Pith},
  title        = {Pith review of: DUN-SRE: Deep Unrolling Network with Spatiotemporal Rotation Equivariance for Dynamic MRI Reconstruction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TZ367Z7C}},
  note         = {Machine review of arXiv:2506.10309}
}
read the original abstract

Dynamic Magnetic Resonance Imaging (MRI) exhibits transformation symmetries, including spatial rotation symmetry within individual frames and temporal symmetry along the time dimension. Explicit incorporation of these symmetry priors in the reconstruction model can significantly improve image quality, especially under aggressive undersampling scenarios. Recently, Equivariant convolutional neural network (ECNN) has shown great promise in exploiting spatial symmetry priors. However, existing ECNNs critically fail to model temporal symmetry, arguably the most universal and informative structural prior in dynamic MRI reconstruction. To tackle this issue, we propose a novel Deep Unrolling Network with Spatiotemporal Rotation Equivariance (DUN-SRE) for Dynamic MRI Reconstruction. The DUN-SRE establishes spatiotemporal equivariance through a (2+1)D equivariant convolutional architecture. In particular, it integrates both the data consistency and proximal mapping module into a unified deep unrolling framework. This architecture ensures rigorous propagation of spatiotemporal rotation symmetry constraints throughout the reconstruction process, enabling more physically accurate modeling of cardiac motion dynamics in cine MRI. In addition, a high-fidelity group filter parameterization mechanism is developed to maintain representation precision while enforcing symmetry constraints. Comprehensive experiments on Cardiac CINE MRI datasets demonstrate that DUN-SRE achieves state-of-the-art performance, particularly in preserving rotation-symmetric structures, offering strong generalization capability to a broad range of dynamic MRI reconstruction tasks.

Figures

Figures reproduced from arXiv: 2506.10309 by the authors.

Figure 1
Figure 1. (a) An MR phantom image with rotation symmetry pairs indicated [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. A group of weight-sharing SREC filters can extract all similar [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Basic structure of an example network constructed using proposed SREC for dynamic MR reconstruction, where we set the transformation group [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: The full pipeline of DUN-SRE for dynamic MR reconstruction. A fully sampled calibration region is extracted from the time-averaged k-space data [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Ablation study of the proposed method DUN-SRE on the in-house acquired cardiac cine dataset. Quantitative results of model variants evaluated [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Qualitative results of ablations under 12× and 20× acceleration on [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Qualitative reconstruction results of comparison study in the spatial domain (x–y) under different acceleration factors. The second row shows the [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 8
Figure 8. Figure 8: Results of comparison study along the temporal dimension (y–t) under [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: Visual comparisons for the generalization study on the OCMR [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Reconstruction results of prospective study on real-time 9× under [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 52 canonical work pages

  1. [1]

    Integrated MRI-guided radiotherapy—opportunities and challenges,

    P. J. Keall, C. Brighi, C. Glide-Hurst, G. Liney, P. Z. Liu, S. Lydiard, C. Paganelli, T. Pham, S. Shan, A. C. Treeet al., “Integrated MRI-guided radiotherapy—opportunities and challenges,”Nature Reviews Clinical Oncology, vol. 19, no. 7, pp. 458–470, 2022

    work page 2022

  2. [2]

    Screening and diagnosis of cardiovascular disease using artificial intelligence-enabled cardiac magnetic resonance imaging,

    Y .-R. Wang, K. Yang, Y . Wen, P. Wang, Y . Hu, Y . Lai, Y . Wang, K. Zhao, S. Tang, A. Zhanget al., “Screening and diagnosis of cardiovascular disease using artificial intelligence-enabled cardiac magnetic resonance imaging,”Nature Medicine, vol. 30, no. 5, pp. 1471–1480, 2024

    work page 2024

  3. [3]

    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: An Official Journal of the International Society for Magnetic Resonance in Medicine, vol. 58, no. 6, pp. 1182–1195, 2007

    2007

  4. [4]

    Model-based image reconstruction for MRI,

    J. A. Fessler, “Model-based image reconstruction for MRI,”IEEE signal processing magazine, vol. 27, no. 4, pp. 81–89, 2010

    work page 2010

  5. [5]

    RARE: Image reconstruction using deep priors learned without groundtruth,

    J. Liu, Y . Sun, C. Eldeniz, W. Gan, H. An, and U. S. Kamilov, “RARE: Image reconstruction using deep priors learned without groundtruth,” IEEE Journal of Selected Topics in Signal Processing, vol. 14, no. 6, pp. 1088–1099, 2020

    work page 2020

  6. [6]

    Optimization methods for magnetic resonance image reconstruction: Key models and optimization algorithms,

    J. A. Fessler, “Optimization methods for magnetic resonance image reconstruction: Key models and optimization algorithms,”IEEE signal processing magazine, vol. 37, no. 1, pp. 33–40, 2020

    work page 2020

  7. [7]

    Image reconstruction from highly undersampled (k, t)-space data with joint partial separability and sparsity constraints,

    B. Zhao, J. P. Haldar, A. G. Christodoulou, and Z.-P. Liang, “Image reconstruction from highly undersampled (k, t)-space data with joint partial separability and sparsity constraints,”IEEE transactions on medical imaging, vol. 31, no. 9, pp. 1809–1820, 2012

    work page 2012

  8. [8]

    Low-rank modeling of localk-space neighborhoods (LO- RAKS) for constrained MRI,

    J. P. Haldar, “Low-rank modeling of localk-space neighborhoods (LO- RAKS) for constrained MRI,”IEEE transactions on medical imaging, vol. 33, no. 3, pp. 668–681, 2013

    work page 2013

Show all 66 references
  1. [9]

    Accelerated high-dimensional MR imaging with sparse sampling using low-rank tensors,

    J. He, Q. Liu, A. G. Christodoulou, C. Ma, F. Lam, and Z.-P. Liang, “Accelerated high-dimensional MR imaging with sparse sampling using low-rank tensors,”IEEE transactions on medical imaging, vol. 35, no. 9, pp. 2119–2129, 2016

    2016

  2. [10]

    Reducing the dimensionality of data with neural networks,

    G. E. Hinton and R. R. Salakhutdinov, “Reducing the dimensionality of data with neural networks,”science, vol. 313, no. 5786, pp. 504–507, 2006

    2006

  3. [11]

    Deep learning,

    Y . LeCun, Y . Bengio, and G. Hinton, “Deep learning,”nature, vol. 521, no. 7553, pp. 436–444, 2015

    2015

  4. [12]

    Artificial intelligence in healthcare,

    K.-H. Yu, A. L. Beam, and I. S. Kohane, “Artificial intelligence in healthcare,”Nature biomedical engineering, vol. 2, no. 10, pp. 719–731, 2018

    2018

  5. [13]

    Deep magnetic resonance image reconstruction: Inverse problems meet neural networks,

    D. Liang, J. Cheng, Z. Ke, and L. Ying, “Deep magnetic resonance image reconstruction: Inverse problems meet neural networks,”IEEE Signal Processing Magazine, vol. 37, no. 1, pp. 141–151, 2020

    2020

  6. [14]

    Dense recurrent neural networks for accelerated MRI: History- cognizant unrolling of optimization algorithms,

    S. A. H. Hosseini, B. Yaman, S. Moeller, M. Hong, and M. Akc ¸akaya, “Dense recurrent neural networks for accelerated MRI: History- cognizant unrolling of optimization algorithms,”IEEE Journal of Se- lected Topics in Signal Processing, vol. 14, no. 6, pp. 1280–1291, 2020

    2020

  7. [15]

    Geometric approaches to increase the expressivity of deep neural networks for MR reconstruction,

    E. Cha, G. Oh, and J. C. Ye, “Geometric approaches to increase the expressivity of deep neural networks for MR reconstruction,”IEEE Journal of Selected Topics in Signal Processing, vol. 14, no. 6, pp. 1292–1305, 2020

    2020

  8. [16]

    High-throughput deep unfolding network for compressive sensing MRI,

    J. Zhang, Z. Zhang, J. Xie, and Y . Zhang, “High-throughput deep unfolding network for compressive sensing MRI,”IEEE Journal of Selected Topics in Signal Processing, vol. 16, no. 4, pp. 750–761, 2022

    2022

  9. [17]

    AI-based reconstruction for fast MRI—A systematic review and meta-analysis,

    Y . Chen, C.-B. Sch ¨onlieb, P. Li `o, T. Leiner, P. L. Dragotti, G. Wang, D. Rueckert, D. Firmin, and G. Yang, “AI-based reconstruction for fast MRI—A systematic review and meta-analysis,”Proceedings of the IEEE, vol. 110, no. 2, pp. 224–245, 2022

    2022

  10. [18]

    Convolutional recurrent neural networks for dynamic MR image reconstruction,

    C. Qin, J. Schlemper, J. Caballero, A. N. Price, J. V . Hajnal, and D. Rueckert, “Convolutional recurrent neural networks for dynamic MR image reconstruction,”IEEE transactions on medical imaging, vol. 38, no. 1, pp. 280–290, 2018

    2018

  11. [19]

    Cinenet: deep learning-based 3d cardiac CINE MRI reconstruction with multi- coil complex-valued 4d spatio-temporal convolutions,

    T. K ¨ustner, N. Fuin, K. Hammernik, A. Bustin, H. Qi, R. Hajhosseiny, P. G. Masci, R. Neji, D. Rueckert, R. M. Botnaret al., “Cinenet: deep learning-based 3d cardiac CINE MRI reconstruction with multi- coil complex-valued 4d spatio-temporal convolutions,”Scientific reports, v...

    2020

  12. [20]

    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

    2020

  13. [21]

    Souza, Y

    R. Souza, Y . Beauferris, W. Loos, R. M. Lebel, and R. Frayne, “En- hanced deep-learning-based magnetic resonance image reconstruction by leveraging prior subject-specific brain imaging: Proof-of-concept using a cohort of presumed normal subjects,”IEEE Journal of Selected Topi...

    2020

  14. [22]

    A deep learning method for optimal undersampling patterns and image recovery for MRI exploiting losses and projections,

    F. Martinini, M. Mangia, A. Marchioni, R. Rovatti, and G. Setti, “A deep learning method for optimal undersampling patterns and image recovery for MRI exploiting losses and projections,”IEEE Journal of Selected Topics in Signal Processing, vol. 16, no. 4, pp. 713–724, 2022

    2022

  15. [23]

    PUERT: Probabilistic under- sampling and explicable reconstruction network for CS-MRI,

    J. Xie, J. Zhang, Y . Zhang, and X. Ji, “PUERT: Probabilistic under- sampling and explicable reconstruction network for CS-MRI,”IEEE Journal of Selected Topics in Signal Processing, vol. 16, no. 4, pp. 737–749, 2022

    2022

  16. [24]

    Going deeper with convolutions,

    C. Szegedy, W. Liu, Y . Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V . Vanhoucke, and A. Rabinovich, “Going deeper with convolutions,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp. 1–9

    2015

  17. [25]

    Group equivariant convolutional networks,

    T. Cohen and M. Welling, “Group equivariant convolutional networks,” inInternational conference on machine learning. PMLR, 2016, pp. 2990–2999

    2016

  18. [26]

    Dense steerable filter CNNs for exploiting rotational symmetry in histology images,

    S. Graham, D. Epstein, and N. Rajpoot, “Dense steerable filter CNNs for exploiting rotational symmetry in histology images,”IEEE Transactions on Medical Imaging, vol. 39, no. 12, pp. 4124–4136, 2020

    2020

  19. [27]

    Roto-translation equivariant convolutional networks: Application to histopathology image analysis,

    M. W. Lafarge, E. J. Bekkers, J. P. Pluim, R. Duits, and M. Veta, “Roto-translation equivariant convolutional networks: Application to histopathology image analysis,”Medical Image Analysis, vol. 68, p. 101849, 2021

    2021

  20. [28]

    General e (2)-equivariant steerable CNNs,

    M. Weiler and G. Cesa, “General e (2)-equivariant steerable CNNs,” Advances in neural information processing systems, vol. 32, 2019

    2019

  21. [29]

    Dicdnet: deep interpretable convolutional dictionary network for metal artifact reduction in CT images,

    H. Wang, Y . Li, N. He, K. Ma, D. Meng, and Y . Zheng, “Dicdnet: deep interpretable convolutional dictionary network for metal artifact reduction in CT images,”IEEE Transactions on Medical Imaging, vol. 41, no. 4, pp. 869–880, 2021

    2021

  22. [30]

    Fourier series expansion based filter parametrization for equivariant convolutions,

    Q. Xie, Q. Zhao, Z. Xu, and D. Meng, “Fourier series expansion based filter parametrization for equivariant convolutions,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 4, pp. 4537– 4551, 2022. IEEE JOURNAL OF SELECTED TOPICS IN SIGNAL PROCESSING 13

    2022

  23. [31]

    Learning steerable filters for rotation equivariant CNNs,

    M. Weiler, F. A. Hamprecht, and M. Storath, “Learning steerable filters for rotation equivariant CNNs,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 849–858

    2018

  24. [32]

    Scale equivariant graph metanetworks,

    I. Kalogeropoulos, G. Bouritsas, and Y . Panagakis, “Scale equivariant graph metanetworks,”Advances in neural information processing sys- tems, vol. 37, pp. 106 800–106 840, 2024

    2024

  25. [33]

    Geometric deep learning and equivariant neural networks,

    J. E. Gerken, J. Aronsson, O. Carlsson, H. Linander, F. Ohlsson, C. Petersson, and D. Persson, “Geometric deep learning and equivariant neural networks,”Artificial Intelligence Review, vol. 56, no. 12, pp. 14 605–14 662, 2023

    2023

  26. [34]

    Imaging with equivariant deep learning: From unrolled network design to fully unsupervised learning,

    D. Chen, M. Davies, M. J. Ehrhardt, C.-B. Sch ¨onlieb, F. Sherry, and J. Tachella, “Imaging with equivariant deep learning: From unrolled network design to fully unsupervised learning,”IEEE Signal Processing Magazine, vol. 40, no. 1, pp. 134–147, 2023

    2023

  27. [35]

    SRE- CNN: A spatiotemporal rotation-equivariant CNN for cardiac cine MR imaging,

    Y . Zhu, J. Cheng, Z.-X. Cui, J. Ren, C. Wang, and D. Liang, “SRE- CNN: A spatiotemporal rotation-equivariant CNN for cardiac cine MR imaging,” inInternational Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 2024, pp. 679–689

    2024

  28. [36]

    Ocmr (v1. 0)–open-access multi-coil k-space dataset for cardiovascular magnetic resonance imaging,

    C. Chen, Y . Liu, P. Schniter, M. Tong, K. Zareba, O. Simonetti, L. Potter, and R. Ahmad, “Ocmr (v1. 0)–open-access multi-coil k-space dataset for cardiovascular magnetic resonance imaging,”arXiv preprint arXiv:2008.03410, 2020

    2008 arXiv

  29. [37]

    Pdo-econvs: Partial differential operator based equivariant convolutions,

    Z. Shen, L. He, Z. Lin, and J. Ma, “Pdo-econvs: Partial differential operator based equivariant convolutions,” inInternational Conference on Machine Learning. PMLR, 2020, pp. 8697–8706

    2020

  30. [38]

    A program to build e (n)-equivariant steerable CNNs,

    G. Cesa, L. Lang, and M. Weiler, “A program to build e (n)-equivariant steerable CNNs,” inInternational conference on learning representa- tions, 2022

    2022

  31. [39]

    Hexaconv,

    E. Hoogeboom, J. W. Peters, T. S. Cohen, and M. Welling, “Hexaconv,” inInternational Conference on Learning Representations, 2018

    2018

  32. [40]

    Harmonic networks: Deep translation and rotation equivariance,

    D. E. Worrall, S. J. Garbin, D. Turmukhambetov, and G. J. Brostow, “Harmonic networks: Deep translation and rotation equivariance,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 5028–5037

    2017

  33. [41]

    Cubenet: Equivariance to 3d rotation and translation,

    D. Worrall and G. Brostow, “Cubenet: Equivariance to 3d rotation and translation,” inProceedings of the European Conference on Computer Vision (ECCV), 2018, pp. 567–584

    2018

  34. [42]

    Spherical CNNs,

    T. S. Cohen, M. Geiger, J. K ¨ohler, and M. Welling, “Spherical CNNs,” inInternational Conference on Learning Representations, 2018

    2018

  35. [43]

    ADMM-csnet: A deep learning approach for image compressive sensing,

    Y . Yang, J. Sun, H. Li, and Z. Xu, “ADMM-csnet: A deep learning approach for image compressive sensing,”IEEE transactions on pattern analysis and machine intelligence, vol. 42, no. 3, pp. 521–538, 2018

    2018

  36. [44]

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

    J. Zhang and B. 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

    2018

  37. [45]

    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

    2018

  38. [46]

    Cascaded dilated dense network with two-step data consistency for MRI reconstruction,

    H. Zheng, F. Fang, and G. Zhang, “Cascaded dilated dense network with two-step data consistency for MRI reconstruction,”Advances in Neural Information Processing Systems, vol. 32, 2019

    2019

  39. [47]

    Transms: Transformers for super-resolution calibration in magnetic particle imaging,

    A. G ¨ung¨or, B. Askin, D. A. Soydan, E. U. Saritas, C. B. Top, and T. C ¸ ukur, “Transms: Transformers for super-resolution calibration in magnetic particle imaging,”IEEE Transactions on Medical Imaging, vol. 41, no. 12, pp. 3562–3574, 2022

    2022

  40. [48]

    High-fidelity MRI reconstruction using adaptive spatial attention selection and deep data consistency prior,

    J. Liu, C. Qin, and M. Yaghoobi, “High-fidelity MRI reconstruction using adaptive spatial attention selection and deep data consistency prior,”IEEE Transactions on Computational Imaging, vol. 9, pp. 298– 313, 2023

    2023

  41. [49]

    DEQ-MPI: A deep equilibrium reconstruction with learned consistency for magnetic particle imaging,

    A. G ¨ung¨or, B. Askin, D. A. Soydan, C. B. Top, E. U. Saritas, and T. C ¸ ukur, “DEQ-MPI: A deep equilibrium reconstruction with learned consistency for magnetic particle imaging,”IEEE Transactions on Med- ical Imaging, vol. 43, no. 1, pp. 321–334, 2023

    2023

  42. [50]

    R. W. Brown, Y .-C. N. Cheng, E. M. Haacke, M. R. Thompson, and R. Venkatesan,Magnetic resonance imaging: physical principles and sequence design. John Wiley & Sons, 2014

    2014

  43. [51]

    Plug-and-play methods for magnetic res- onance imaging: Using denoisers for image recovery,

    R. Ahmad, C. A. Bouman, G. T. Buzzard, S. Chan, S. Liu, E. T. Reehorst, and P. Schniter, “Plug-and-play methods for magnetic res- onance imaging: Using denoisers for image recovery,”IEEE signal processing magazine, vol. 37, no. 1, pp. 105–116, 2020

    2020

  44. [52]

    Learning data consistency and its application to dynamic MR imaging,

    J. Cheng, Z.-X. Cui, W. Huang, Z. Ke, L. Ying, H. Wang, Y . Zhu, and D. Liang, “Learning data consistency and its application to dynamic MR imaging,”IEEE transactions on medical imaging, vol. 40, no. 11, pp. 3140–3153, 2021

    2021

  45. [53]

    Equivariant neural networks for inverse problems,

    E. Celledoni, M. J. Ehrhardt, C. Etmann, B. Owren, C.-B. Sch ¨onlieb, and F. Sherry, “Equivariant neural networks for inverse problems,”Inverse Problems, vol. 37, no. 8, p. 085006, 2021

    2021

  46. [54]

    Scale-equivariant unrolled neural networks for data-efficient accelerated MRI reconstruction,

    B. Gunel, A. Sahiner, A. D. Desai, A. S. Chaudhari, S. Vasanawala, M. Pilanci, and J. Pauly, “Scale-equivariant unrolled neural networks for data-efficient accelerated MRI reconstruction,” inInternational Confer- ence on Medical Image Computing and Computer-Assisted Interventi...

    2022

  47. [55]

    Rotation equivariant proximal operator for deep unfolding methods in image restoration,

    J. Fu, Q. Xie, D. Meng, and Z. Xu, “Rotation equivariant proximal operator for deep unfolding methods in image restoration,”IEEE Trans- actions on Pattern Analysis and Machine Intelligence, 2024

    2024

  48. [56]

    Proximal algorithms,

    N. Parikh, S. Boydet al., “Proximal algorithms,”Foundations and trends® in Optimization, vol. 1, no. 3, pp. 127–239, 2014

    2014

  49. [57]

    A closer look at spatiotemporal convolutions for action recognition,

    D. Tran, H. Wang, L. Torresani, J. Ray, Y . LeCun, and M. Paluri, “A closer look at spatiotemporal convolutions for action recognition,” in Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, 2018, pp. 6450–6459

    2018

  50. [58]

    Vari- able density incoherent spatiotemporal acquisition (VISTA) for highly accelerated cardiac mri,

    R. Ahmad, H. Xue, S. Giri, Y . Ding, J. Craft, and O. P. Simonetti, “Vari- able density incoherent spatiotemporal acquisition (VISTA) for highly accelerated cardiac mri,”Magnetic resonance in medicine, vol. 74, no. 5, pp. 1266–1278, 2015

    2015

  51. [59]

    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

    2014

  52. [60]

    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

    2004

  53. [61]

    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 medical imaging, vol. 30, no. 5, pp. 1028–1041, 2010

    2010

  54. [62]

    Low-rank plus sparse matrix decomposition for accelerated dynamic MRI with separation of background and dynamic components,

    R. Otazo, E. Candes, and D. K. Sodickson, “Low-rank plus sparse matrix decomposition for accelerated dynamic MRI with separation of background and dynamic components,”Magnetic resonance in medicine, vol. 73, no. 3, pp. 1125–1136, 2015

    2015

  55. [63]

    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 medical imaging, vol. 38, no. 2, pp. 394–405, 2018

    2018

  56. [64]

    Accelerating cardiac cine MRI using a deep learning-based espirit reconstruction,

    C. M. Sandino, P. Lai, S. S. Vasanawala, and J. Y . Cheng, “Accelerating cardiac cine MRI using a deep learning-based espirit reconstruction,” Magnetic Resonance in Medicine, vol. 85, no. 1, pp. 152–167, 2021

    2021

  57. [65]

    Deep low-rank plus sparse network for dynamic MR imaging,

    W. Huang, Z. Ke, Z.-X. Cui, J. Cheng, Z. Qiu, S. Jia, L. Ying, Y . Zhu, and D. Liang, “Deep low-rank plus sparse network for dynamic MR imaging,”Medical Image Analysis, vol. 73, p. 102190, 2021

    2021

  58. [66]

    Unrolled and rapid motion-compensated reconstruction for cardiac cine MRI,

    J. Pan, M. Hamdi, W. Huang, K. Hammernik, T. Kuestner, and D. Rueck- ert, “Unrolled and rapid motion-compensated reconstruction for cardiac cine MRI,”Medical Image Analysis, vol. 91, p. 103017, 2024

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.