Pith. sign in

REVIEW 4 major objections 5 minor 30 references

LANTERN: learn analysis transform network for dynamic magnetic resonance imaging with small dataset

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read LANTERN, a deep network that unrolls ADMM with DCT and TV priors, reconstructs dynamic cardiac MRI better than k-t SLR and D5C5 using only 100 training volumes.

desk verdict A plausibly useful unrolled ADMM network for dynamic MRI, but the central small-data claim is asserted rather than demonstrated because D5C5 is never trained on the same 100 volumes. read the letter →

arxiv 1908.09140 v1 pith:UA6DEEAN submitted 2019-08-24 eess.IV cs.CV

Shanshan Wang , Yanxia Chen , Taohui Xiao , Ziwen Ke , Qiegen Liu , Hairong Zheng This is my paper · ORCID
classification eess.IVcs.CV
keywords dynamicMRIreconstructioncardiaccinecompressedsensingADMMunrollingdeeplearningsmalltrainingdatasetk-spaceundersamplinganalysistransform
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 a model-based deep network, LANTERN, can reconstruct dynamic cardiac MRI from heavily undersampled k-space data using only 100 training volumes, while outperforming both the compressed-sensing baseline k-t SLR and the large-data deep baseline D5C5. The significance is that dynamic MRI acceleration currently faces a trade-off: classical compressed sensing needs hand-tuned parameters, while deep reconstruction methods typically require thousands of training samples. LANTERN unrolls an ADMM solver of a sparse convolutional coding model into network layers and learns all parameters end-to-end. On 1D random (2x-11x) and 2D radial (2x-15x) undersampling, it reports higher PSNR, SSIM, and HFEN than both baselines, with reconstruction in under three seconds. If correct, this would make high-quality accelerated dynamic MRI practical in settings where large training datasets are unavailable.

What carries the argument

The load-bearing object is the unrolled ADMM iteration, expressed as six layer types: Recon, Addition, Conv1, Nonlinear, Conv2, and Multi. Conv1 and Conv2 learn convolutional analysis operators that sparsify the image, the Nonlinear layer approximates the derivative of the regularization function by a piecewise linear function, and the Recon layer performs the closed-form k-space data-consistency update. The DCT and TV initialization provides the prior structure that lets the network work from a small training set.

What would settle it

Train D5C5 (or an equivalent data-hungry deep reconstruction network) on the same 100 training volumes used for LANTERN, with identical test data and sampling masks, and compare PSNR, SSIM, and HFEN; if the retrained D5C5 matches or beats LANTERN's reported 41.44 dB PSNR at 1D random 4x acceleration, the central small-data advantage claim fails.

Watch

Extended reading notes

Core claim

LANTERN treats dynamic MR reconstruction as a sparse convolutional coding problem with a data-fidelity term and learned analysis transforms, and solves it by unrolling alternating direction method of multipliers (ADMM) iterations into a fixed-depth network. The spatial transform is initialized with the discrete cosine transform (DCT) and the temporal transform with total variation (TV), and every parameter—filters, regularization weights, step sizes, and the piecewise-linear nonlinearity—is learned end-to-end from undersampled-to-fully-sampled pairs. On cardiac data, the trained network reaches higher average PSNR, SSIM, and HFEN than k-t SLR and D5C5 at every tested acceleration factor, including 11x random and 15x radial undersampling, while training on 100 volumes versus 2900 for D5C5.

Load-bearing premise

The comparison assumes D5C5 is a valid control at the same training size, but the paper trains D5C5 on 2900 volumes and LANTERN on 100, asserting without shown evidence that D5C5 performs poorly with 100 volumes.

Editorial extensions

If this is right

  • If the central claim holds, dynamic cardiac MRI can be reconstructed accurately from 2x-11x random or 2x-15x radial undersampling with a network trained on 100 volumes, far fewer than the thousands used by the D5C5 baseline.
  • The unrolled ADMM structure preserves a compressed-sensing interpretation: the learned filters act as data-adaptive sparsifying transforms, and the data-fidelity update remains explicit.
  • Reconstruction time under three seconds, versus about 200 seconds for k-t SLR, would make the approach clinically practical despite a training phase measured in hours.
  • Because all regularization and step-size parameters are learned, the manual tuning burden of classical CS-MRI reconstruction is removed.

Reading between the lines

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

  • A decisive test the paper does not run is to train D5C5 on the same 100 volumes; if that retrained network also reconstructs well, the claimed small-data advantage would be an artifact of comparison size rather than of the LANTERN architecture.
  • The DCT+TV initialization likely acts as an inductive bias that reduces data demand; ablating the initializer while keeping the rest of the architecture fixed would isolate how much of the small-data performance comes from initialization versus learned layers.
  • The same recipe—unrolling an ADMM solver and initializing analytic transforms—could transfer to other inverse problems with known sparsifying bases, such as CT or PET, but the paper does not test this.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes LANTERN, an unrolled ADMM network for dynamic MRI reconstruction that learns analysis sparsifying transforms via a small set of training volumes. The network is initialized with DCT (spatial) and TV (temporal) filters, unrolls a fixed number of ADMM stages, and replaces the derivative of the prior with a learned piecewise-linear function. On 1D random and 2D radial undersampling masks, the authors report higher PSNR, SSIM, and HFEN than k-t SLR and D5C5 across acceleration factors from 2x to 15x, training with 100 cardiac volumes. The paper claims that this demonstrates superior reconstruction accuracy when the amount of training data is limited.

Significance. If the small-data claim is established, the result would be practically significant for clinical MRI, where large fully-sampled training sets are often unavailable. The unrolled architecture is a credible model-based deep learning approach, and the paper provides useful details: quantitative metrics with standard deviations, a comparison of initialization schemes, and training/validation loss curves. The attempt to combine a learnable analysis prior with DCT/TV initialization is a reasonable design choice. However, the central claim is currently supported only by an indirect comparison: the deep-learning baseline D5C5 is trained on 2900 volumes, not on the same 100 volumes used for LANTERN, and the reported LANTERN numbers are partially selected on the test set. The significance is therefore conditional on a matched evaluation.

major comments (4)
  1. [Implementation (paragraph on D5C5); Tables 4 and 5] The paper's core claim that LANTERN is superior to deep-learning baselines under small training data is not directly tested. D5C5 is trained on 2900 volumes (300 testing) while LANTERN is trained on 100 volumes and tested on 50. The text asserts that with 100 training data 'the reconstruction results are not good enough' and that experiments `verify the conjecture', but no experiment with D5C5 trained on the same 100 volumes is reported or shown. As written, the comparison conflates architecture advantage with training-set-size advantage; if D5C5 also performs well at 100 training volumes, the claimed small-data advantage is not established. Reporting this matched control is essential.
  2. [Impact of dataset size (Table 2, Fig. 3); Effect of initializations (Table 3)] The training-set size and the initialization scheme are selected using the 50 test volumes. Table 2 reports reconstruction metrics on the test set for training sizes 50, 60, 80, 100, 120, and the text says 'we selected 100 images' because the test PSNR is best at 100. Similarly, Table 3 compares Random Gaussian, DCT, and DCT+TV initializations on the same 50 test volumes and chooses DCT+TV. This makes the headline numbers in Tables 4 and 5 partially test-selected, so they are not a clean held-out evaluation. A validation split (or nested selection) should be used for these choices, and the final test numbers should be reported after the selection is fixed.
  3. [Dataset paragraph and Implementation paragraph] The data description is internally inconsistent and matters for the small-data claim. The Dataset section says 101 fully sampled cardiac datasets were collected and that '150 cardiac data were generated with 100 data for network training, and 50 for network testing', while the Implementation section says the data were cut to 126x126x16 and 3200 data were obtained, with 2900 for D5C5 training and 300 for testing. It is unclear how 101 datasets become 150, how many distinct subjects or slices are in the 100/50 split, and whether the 2900 D5C5 volumes overlap with the 100 LANTERN volumes. Please clarify this pipeline and, ideally, evaluate LANTERN on the same 300-volume test set used for D5C5, or evaluate D5C5 on the same 50-volume test set.
  4. [Back-propagation, ReCon layer gradient (paragraph after Eq. (8))] The expression for d x^(n) / d rho^(n) appears to use a scalar-style derivative of a matrix inverse, writing terms like (P^H P + rho I)^{-2} with the inverse raised to a power. As written this is not a valid matrix derivative. Since the end-to-end training of the network is a central contribution, the gradient derivation for the rho parameters should be either corrected, stated with proper operator notation (e.g., - (P^H P + rho I)^{-1} (d/dt)[P^H P + rho I] (P^H P + rho I)^{-1}), or justified by reference to a standard matrix calculus result.
minor comments (5)
  1. [Loss definition] The text says the loss is the standard mean square error, but the equation for E(Theta) defines a normalized RMSE loss (square root of a normalized sum of squared errors). Please align the text with the actual loss.
  2. [Abstract] There is a typo, 'undersamling', in the abstract.
  3. [Eq. (8) and layer notation] The notation in Eqs. (7)-(8) is inconsistent: the Addition layer uses v^(n,k-1) while the multiplier update uses v^(n), and the warm-start cases for n=1 are written separately but not clearly tied to the general formulas. Please unify the superscripts and state explicitly which v is passed to the Multi layer.
  4. [Filter description] The sentence 'the size of the applied filter is 3*3*9, where the first eight (3*3*8) are DCT and the last one (3*3*1) is TV' is unclear: please specify how the temporal dimension is arranged and how the 3D convolution and 2D TV filter are combined.
  5. [Convergence analysis (Fig. 10)] The convergence analysis only shows that training and validation losses decrease. This is evidence against overfitting, but it does not by itself demonstrate that the unrolled network converges to a solution of the original optimization problem. Consider adding a plot of reconstruction error on a fixed validation set per epoch, or a comparison with the optimization-based solution at initialization.

Circularity Check

3 steps flagged · score 6.0 of 10

Central small-data advantage is partially circular: test-set metrics are used to choose training size and initialization, and D5C5 is never tested on the same 100 volumes.

  1. fitted input called prediction [Section 'Impact of dataset size', Table 2 and Fig. 3]
    "And from the results, it can be found that the reconstruction result is the best when the data amount is 100. Therefore, we selected 100 images for experiments in subsequent works."

    The number of training volumes is treated as a free design choice and is selected by comparing PSNR, SSIM, and HFEN on the same 50 test volumes (Table 2). The configuration with the best test metrics, N=100, is then fixed for all later comparisons in Tables 4-5 and Figures 5-9. Consequently the reported LANTERN test performance is not a held-out prediction: the test set was used to choose the training-set size, so the final numbers are by construction at least as favorable as the alternatives that were explicitly compared. The generalization claim is therefore partly fitted to the evaluation data.

  2. fitted input called prediction [Section 'Effect of initializations: Random Gauss, DCT, DCT+TV', Table 3]
    "The PSNR value of DCT+TV is also the highest. Table 3 is the average quantization index corresponding to the reconstruction result under the data volume of 100 and 1Drandom 4x acceleration, which also proves that the DCT+TV method is superior to the other two initialization methods."

    The initialization scheme is selected by evaluating Random Gauss, DCT, and DCT+TV on the same 50 test volumes (Table 3) and choosing the variant with the highest PSNR. That test-selected initialization is then used in every subsequent experiment and comparison. Thus an important component of LANTERN's final test metrics is fitted to the test set rather than validated on independent data, making the reported superiority of DCT+TV self-confirming under the paper's own protocol.

1 more flagged steps
  1. other [Section 'Implementation']
    "Considering that D5C5 is a method based on big data sets, if we use 100 data to train, the reconstruction results are not good enough, and the experimental results also verify the conjecture. So, for the D5C5 method, we cut our data to 126*126*16 and get 3200 data, 2900 data for training, 300 data for testing"

    The paper's central claim is that LANTERN is more accurate than deep-learning methods when training data are limited. The only deep-learning comparator, D5C5, is never trained on the same 100 volumes; instead it is trained on 2900 volumes, and the paper asserts without presenting the experiment that 100-volume D5C5 'results are not good enough' and that 'experimental results also verify the conjecture.' The advantage over D5C5 is therefore not measured in the claimed regime: the comparison protocol assumes the comparator's small-data failure, so the observed outcome is embedded in the experimental design rather than independently demonstrated.

full rationale

The mathematical derivation in Eqs. (1)-(8) is self-contained: the ADMM unrolling, the learned convolutional filters, and the SPLF nonlinearity are not circular, and the architecture is credited to the external ADMM-Net [23]. The self-citation [21] supports only a background statement and is not load-bearing. The circularity is in the evaluation protocol. First, the training-set size is chosen by test-set PSNR (Table 2, Fig. 3): the paper states the result is best at 100 and then selects 100 for all subsequent experiments, so the reported test metrics are partially fitted to the test set. Second, the DCT+TV initialization is chosen on the same test set (Table 3), adding another test-selected component to the final numbers. Third, the central small-data advantage over D5C5 is asserted rather than demonstrated: D5C5 is never trained on the same 100 volumes, and the paper's own Implementation section says 100-volume D5C5 would not be good enough 'and the experimental results also verify the conjecture,' without presenting such an experiment. The comparison therefore embeds the desired outcome in the protocol. Because the core derivation itself is independent but the headline claim is partly manufactured by test-set selection and an unequal control, the score is 6.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The network weights are trained, not free parameters, but the evaluation pipeline selects the training-set size and initialization from the test set, and the hand-set ADMM/architecture hyperparameters are not ablated. The core assumptions are standard MRI physics plus design choices common to unrolled networks.

free parameters (4)
  • Training data amount = 100 volumes
    Selected as optimal based on test-set PSNR in Table 2 and Figure 3; 120 volumes gave slightly lower SSIM/HFEN.
  • Filter initialization scheme = DCT+TV
    Selected by comparing Random Gaussian, DCT, and DCT+TV initializations on the test set (Table 3).
  • Step-size initializations = rho=0.2, lr=0.3, eta~=1.8
    Hand-set initial values for the unrolled ADMM; these are learned during end-to-end training, so the initial values are arbitrary but the reported performance depends on the chosen starting point.
  • Network depth and filter size = stage=13, substage=1, filter size 3x3x9
    Hand-chosen architecture hyperparameters, not justified by ablation or analysis.
assumptions (4)
  • domain assumption The measured k-space data is related to the image by a linear Fourier measurement operator with additive noise (Eq. 1).
    Standard model for MRI acquisition; the paper does not validate residual noise or model mismatch.
  • domain assumption The dynamic image sequence is sparsely representable by a learnable analysis transform with spatial DCT-like and temporal TV-like filters.
    This is the prior that the network learns; it is plausible for cardiac cine MRI but not proven beyond the reported experiments.
  • ad hoc to paper The derivative of the prior regularization function can be approximated by a learned piecewise-linear function SPLF.
    Introduced in the Nonlinear layer; this is a network design choice that replaces the analytic proximal/derivative step in ADMM, and its adequacy is not theoretically analyzed.
  • ad hoc to paper Unrolling K=1 inner ADMM step per stage and training end-to-end yields a valid approximation of the original optimization.
    The paper sets substage=1 and does not provide convergence guarantees for the unrolled network; standard practice in ADMM-Net but not justified here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LANTERN: learn analysis transform network for dynamic magnetic resonance imaging with small dataset." pith.science (2026). https://pith.science/paper/UA6DEEAN

@misc{pith2026190809140,
  author       = {Pith},
  title        = {Pith review of: LANTERN: learn analysis transform network for dynamic magnetic resonance imaging with small dataset},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UA6DEEAN}},
  note         = {Machine review of arXiv:1908.09140}
}
read the original abstract

This paper proposes to learn analysis transform network for dynamic magnetic resonance imaging (LANTERN) with small dataset. Integrating the strength of CS-MRI and deep learning, the proposed framework is highlighted in three components: (i) The spatial and temporal domains are sparsely constrained by using adaptively trained CNN. (ii) We introduce an end-to-end framework to learn the parameters in LANTERN to solve the difficulty of parameter selection in traditional methods. (iii) Compared to existing deep learning reconstruction methods, our reconstruction accuracy is better when the amount of data is limited. Our model is able to fully exploit the redundancy in spatial and temporal of dynamic MR images. We performed quantitative and qualitative analysis of cardiac datasets at different acceleration factors (2x-11x) and different undersampling modes. In comparison with state-of-the-art methods, extensive experiments show that our method achieves consistent better reconstruction performance on the MRI reconstruction in terms of three quantitative metrics (PSNR, SSIM and HFEN) under different undersamling patterns and acceleration factors.

Figures

Figures reproduced from arXiv: 1908.09140 by the authors.

Figure 1
Figure 1. The proposed LANTERN network architecture for dMRI reconstruction. In (A) and (B), the blue arrow indicates the process of reconstructing the undersampled image by forward propagation. The pink arrow indicates the process of back-propagation the update parameters [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

  1. [1]

    Simultaneous acquisition of spatial harmonics (SMASH): Fast imaging with radiofrequency coil arrays,

    D. K. Sodickson and W. J. Manning, “Simultaneous acquisition of spatial harmonics (SMASH): Fast imaging with radiofrequency coil arrays,” Magn. Reson. Med., vol. 38, no. 4, pp. 591–603, 1997

    work page 1997

  2. [2]

    Mr imaging of the breast: Fast imaging sequences with and without Gd-DTPA. Preliminary observations,

    W. A. Kaiser and E. Zeitler, “Mr imaging of the breast: Fast imaging sequences with and without Gd-DTPA. Preliminary observations,” Radiology, vol. 170, no. 3, pp. 681–686, 1989

    work page 1989

  3. [3]

    Dictionary learning and time sparsity for dynamic MR data reconstruction,

    J. Caballero, A. N. Price, D. Rueckert, and J. V Hajnal, “Dictionary learning and time sparsity for dynamic MR data reconstruction,” IEEE Trans. Med. Imaging, vol. 33, no. 4, pp. 979–994, 2014

    work page 2014

  4. [4]

    A deep cascade of convolutional neural networks for dynamic MR image reconstruction,

    J. Schlemper, J. Caballero, J. V Hajnal, A. N. Price, and D. Rueckert, “A deep cascade of convolutional neural networks for dynamic MR image reconstruction,” IEEE Trans. Med. Imaging, vol. 37, no. 2, pp. 491–503, 2017

    work page 2017

  5. [5]

    k‐t FOCUSS: a general compressed sensing framework for high resolution dynamic MRI,

    H. Jung, K. Sung, K. S. Nayak, E. Y. Kim, and J. C. Ye, “k‐t FOCUSS: a general compressed sensing framework for high resolution dynamic MRI,” Magn. Reson. Med. An Off. J. Int. Soc. Magn. Reson. Med., vol. 61, no. 1, pp. 103–116, 2009

    work page 2009

  6. [6]

    k-t BLAST and k-t SENSE: Dynamic MRI With High Frame Rate Exploiting Spatiotemporal Correlations,

    J. Tsao, P. Boesiger, and K. P. Pruessmann, “k-t BLAST and k-t SENSE: Dynamic MRI With High Frame Rate Exploiting Spatiotemporal Correlations,” Magn. Reson. Med., vol. 50, no. 5, pp. 1031–1042, 2003

    work page 2003

  7. [7]

    K-t ISD: Dynamic cardiac MR imaging using compressed sensing with iterative support detection,

    D. Liang, E. V. R. DiBella, R. R. Chen, and L. Ying, “K-t ISD: Dynamic cardiac MR imaging using compressed sensing with iterative support detection,” Magn. Reson. Med., vol. 68, no. 1, pp. 41–53, 2012

    work page 2012

  8. [8]

    Undersampled radial MRI with multiple coils. Iterative image reconstruction using a total variation constraint,

    K. T. Block, M. Uecker, and J. Frahm, “Undersampled radial MRI with multiple coils. Iterative image reconstruction using a total variation constraint,” Magn. Reson. Med., vol. 57, no. 6, pp. 1086–1098, 2007

    work page 2007

Show all 30 references
  1. [9]

    An iterative regularization method for total variation-based image restoration,

    S. Osher, M. Burger, D. Goldfarb, J. Xu, and W. Yin, “An iterative regularization method for total variation-based image restoration,” Multiscale Model. Simul., vol. 4, no. 2, pp. 460–489, 2005

    2005

  2. [10]

    A wavelet-based regularized reconstruction algorithm for SENSE parallel MRI with applications to neuroimaging,

    L. Chaâri, J. C. Pesquet, A. Benazza-Benyahia, and P. Ciuciu, “A wavelet-based regularized reconstruction algorithm for SENSE parallel MRI with applications to neuroimaging,” Med. Image Anal., vol. 15, no. 2, pp. 185–201, 2011

    2011

  3. [11]

    Improved dynamic MRI reconstruction by exploiting sparsity and rank- deficiency,

    A. Majumdar, “Improved dynamic MRI reconstruction by exploiting sparsity and rank- deficiency,” Magn. Reson. Imaging, vol. 31, no. 5, pp. 789–795, 2013

    2013

  4. [12]

    Non-convex algorithm for sparse and low-rank recovery: Application to dynamic MRI reconstruction,

    A. Majumdar, R. K. Ward, and T. Aboulnasr, “Non-convex algorithm for sparse and low-rank recovery: Application to dynamic MRI reconstruction,” Magn. Reson. Imaging, vol. 31, no. 3, pp. 448–455, 2013

    2013

  5. [13]

    Accelerated dynamic MRI exploiting sparsity and low-rank structure: K-t SLR,

    S. G. Lingala, Y. Hu, E. Dibella, and M. Jacob, “Accelerated dynamic MRI exploiting sparsity and low-rank structure: K-t SLR,” IEEE Trans. Med. Imaging, 2011

    2011

  6. [14]

    Generalized kt BLAST and kt SENSE using FOCUSS,

    H. Jung, J. Yoo, and J. C. Ye, “Generalized kt BLAST and kt SENSE using FOCUSS,” in 2007 4th IEEE International Symposium on Biomedical Imaging: From Nano to Macro , 2007, pp. 145– 148

    2007

  7. [15]

    Motion estimated and compensated compressed sensing dynamic magnetic resonance imaging: What we can learn from video compression techniques,

    H. Jung and J. C. Ye, “Motion estimated and compensated compressed sensing dynamic magnetic resonance imaging: What we can learn from video compression techniques,” Int. J. Imaging Syst. Technol., vol. 20, no. 2, pp. 81–98, 2010

    2010

  8. [16]

    Improving synthesis and analysis prior blind compressed sensing with low -rank constraints for dynamic MRI reconstruction,

    A. Majumdar, “Improving synthesis and analysis prior blind compressed sensing with low -rank constraints for dynamic MRI reconstruction,” Magn. Reson. Imaging, vol. 33, no. 1, pp. 174– 179, 2015

    2015

  9. [17]

    KT sparse: high frame -rate dynamic magnetic resonance imaging exploiting spatio-temporal sparsity

    M. Lustig, J. M. Santos, D. L. Donoho, and J. M. Pauly, “KT sparse: high frame -rate dynamic magnetic resonance imaging exploiting spatio-temporal sparsity.” Google Patents, 2009

    2009

  10. [18]

    Undersampled dynamic magnetic resonance imaging using patch-based spatiotemporal dictionaries,

    Y. Wang, Y. Zhou, and L. Ying, “Undersampled dynamic magnetic resonance imaging using patch-based spatiotemporal dictionaries,” in 2013 IEEE 10th International Symposium on Biomedical Imaging, 2013, pp. 294–297

    2013

  11. [19]

    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 Trans. Med. Imaging, vol. 31, no. 9, pp. 1809–1820, 2012

    2012

  12. [20]

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

    R. Otazo, E. Candès, and D. K. Sodickson, “Low-rank plus sparse matrix decomposition for accelerated dynamic MRI with separation of background and dynamic components,” Magn. Reson. Med., 2015

    2015

  13. [21]

    Dimension: Dynamic mr imaging with both k-space and spatial prior knowledge obtained via multi-supervised network training,

    S. Wang, Z. Ke, H. Cheng, S. Jia, Y. Leslie, H. Zheng, and D. Liang, “Dimension: Dynamic mr imaging with both k-space and spatial prior knowledge obtained via multi-supervised network training,” arXiv Prepr. arXiv1810.00302, 2018

    arXiv

  14. [22]

    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,” Magn. Reson. Med., 2018

    2018

  15. [23]

    Deep ADMM-Net for compressive sensing MRI,

    J. Sun, H. Li, and Z. Xu, “Deep ADMM-Net for compressive sensing MRI,” in Advances in neural information processing systems, 2016, pp. 10–18

    2016

  16. [24]

    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 Trans. Med. Imaging, vol. 38, no. 2, pp. 394–405, 2018

    2018

  17. [25]

    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, p. 487, 2018

    2018

  18. [26]

    Deep learning with domain adaptation for accelerated projection‐reconstruction MR,

    Y. Han, J. Yoo, H. H. Kim, H. J. Shin, K. Sung, and J. C. Ye, “Deep learning with domain adaptation for accelerated projection‐reconstruction MR,” Magn. Reson. Med., vol. 80, no. 3, pp. 1189–1205, 2018

    2018

  19. [27]

    KIKI‐net: cross‐domain convolutional neural networks for reconstructing undersampled magnetic resonance images,

    T. Eo, Y. Jun, T. Kim, J. Jang, H. Lee, and D. Hwang, “KIKI‐net: cross‐domain convolutional neural networks for reconstructing undersampled magnetic resonance images,” Magn. Reson. Med., vol. 80, no. 5, pp. 2188–2201, 2018

    2018

  20. [28]

    Compressed sensing MRI using a recursive dilated network,

    L. Sun, Z. Fan, Y. Huang, X. Ding, and J. Paisley, “Compressed sensing MRI using a recursive dilated network,” in Thirty-Second AAAI Conference on Artificial Intelligence, 2018

    2018

  21. [29]

    Compressed sensing MRI reconstruction using a generative adversarial network with a cyclic loss,

    T. M. Quan, T. Nguyen-Duc, and W.-K. Jeong, “Compressed sensing MRI reconstruction using a generative adversarial network with a cyclic loss,” IEEE Trans. Med. Imaging, vol. 37, no. 6, pp. 1488–1497, 2018

    2018

  22. [30]

    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 Trans. Med. Imaging, vol. 38, no. 1, pp. 280–290, 2018

    2018

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.