pith. sign in

arxiv: 1907.09425 · v1 · pith:DFO2ZXHKnew · submitted 2019-07-22 · 📡 eess.IV · cs.CV

k-t NEXT: Dynamic MR Image Reconstruction Exploiting Spatio-temporal Correlations

Chen Qin , Jo Schlemper , Jinming Duan , Gavin Seegoolam , Anthony Price , Joseph Hajnal , Daniel Rueckert This is my paper

Pith reviewed 2026-05-24 17:40 UTC · model grok-4.3

classification 📡 eess.IV cs.CV
keywords dynamic MRIimage reconstructiondeep learningspatio-temporal correlationsx-f domaincardiac cine MRIundersampled datak-t space
0
0 comments X

The pith

k-t NEXT reconstructs dynamic MR images from undersampled data by alternating between x-f and image domains to exploit spatio-temporal correlations.

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

Dynamic MRI exhibits high correlations in k-space and time that can be used to accelerate imaging from undersampled acquisitions. The paper proposes k-t NEXT, a deep learning method inspired by traditional k-t BLAST and k-t FOCUSS approaches that first reconstructs signals in the x-f domain. The network then learns to recover the signals by alternating the reconstruction process between x-f space and image space in an iterative fashion. This enables the network to capture useful information and jointly exploit spatio-temporal correlations from both complementary domains. Experiments on highly undersampled short-axis cardiac cine MRI scans show that the method outperforms current state-of-the-art dynamic MR reconstruction approaches both quantitatively and qualitatively.

Core claim

The authors claim that by reconstructing true signals from aliased signals in the x-f domain and then alternating this process with image space reconstruction in an iterative fashion, the k-t NEXT network can effectively capture useful information and jointly exploit spatio-temporal correlations from both complementary domains, resulting in improved dynamic MR image reconstruction from highly undersampled data.

What carries the argument

The iterative alternation of reconstruction between x-f space and image space within the k-t NEXT network, which allows joint exploitation of spatio-temporal correlations from complementary domains.

Load-bearing premise

That alternating reconstruction between x-f space and image space in an iterative fashion enables the network to effectively capture useful information and jointly exploit spatio-temporal correlations from both domains.

What would settle it

A head-to-head test on the same highly undersampled short-axis cardiac cine MRI dataset where k-t NEXT produces no improvement in quantitative metrics such as PSNR or SSIM and no visible qualitative gain over the leading existing method.

Figures

Figures reproduced from arXiv: 1907.09425 by Anthony Price, Chen Qin, Daniel Rueckert, Gavin Seegoolam, Jinming Duan, Jo Schlemper, Joseph Hajnal.

Figure 1
Figure 1. Figure 1: The k-t NEXT reconstruction diagram. True signals can be recovered by iteratively updating the reconstruction in both (a) x-f and (b) image domains via learning the xf-CNN and CRNN jointly. For mathmetical notations, please refer to Eq. 4. where σ (n) rec ∈ C D denotes the complex-valued reconstructed image sequence at iteration n, and σ (0) rec = σu is the acquired zero-filled undersampled images. Here D … view at source ↗
Figure 2
Figure 2. Figure 2: Comparison results on spatial and temporal dimensions with their error [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Visualisation in x-f domain. (a) Ground Truth (b) k-t sampling pattern (c) 9× undersampled data (d) Reconstructed x-f image (e) Error between (c) and (d). Acknowledgements This work was supported by EPSRC programme grant SmartHeart (EP/P001009/1). References 1. Ak¸cakaya, M., Moeller, S., Weing¨artner, S., et al.: Scan-specific robust artificial￾neural-networks for k-space interpolation (RAKI) reconstructi… view at source ↗
read the original abstract

Dynamic magnetic resonance imaging (MRI) exhibits high correlations in k-space and time. In order to accelerate the dynamic MR imaging and to exploit k-t correlations from highly undersampled data, here we propose a novel deep learning based approach for dynamic MR image reconstruction, termed k-t NEXT (k-t NEtwork with X-f Transform). In particular, inspired by traditional methods such as k-t BLAST and k-t FOCUSS, we propose to reconstruct the true signals from aliased signals in x-f domain to exploit the spatio-temporal redundancies. Building on that, the proposed method then learns to recover the signals by alternating the reconstruction process between the x-f space and image space in an iterative fashion. This enables the network to effectively capture useful information and jointly exploit spatio-temporal correlations from both complementary domains. Experiments conducted on highly undersampled short-axis cardiac cine MRI scans demonstrate that our proposed method outperforms the current state-of-the-art dynamic MR reconstruction approaches both quantitatively and qualitatively.

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

0 major / 2 minor

Summary. The manuscript proposes k-t NEXT, a deep learning method for dynamic MR image reconstruction from highly undersampled data. Building on k-t BLAST and k-t FOCUSS, it reconstructs aliased signals in the x-f domain and alternates iteratively between x-f space and image space to jointly exploit spatio-temporal correlations. Experiments on short-axis cardiac cine MRI demonstrate quantitative and qualitative outperformance over current state-of-the-art dynamic MR reconstruction methods.

Significance. If the results hold, the work offers a practical advance in accelerated dynamic MRI by integrating traditional k-t sparsity ideas with learned alternating reconstruction across complementary domains. The manuscript supplies architecture details, training procedure, quantitative comparisons with baselines, and reproducible elements that directly support the reported gains; these are strengths under the journal's standards.

minor comments (2)
  1. [Abstract] Abstract: the claim of outperformance would be strengthened by naming the specific baselines and metrics (e.g., PSNR, SSIM) already used in the results section.
  2. Figure captions: ensure all quantitative panels explicitly reference the corresponding table or section for the reported values.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment of our manuscript and the recommendation to accept. The referee's summary correctly captures the core idea of k-t NEXT as an iterative alternating reconstruction between x-f and image domains to exploit spatio-temporal correlations, building on k-t BLAST and k-t FOCUSS.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper introduces a deep learning network (k-t NEXT) for dynamic MRI reconstruction that alternates between x-f and image domains to learn spatio-temporal correlations. No derivation chain, uniqueness theorem, or fitted parameter is presented that reduces by construction to its own inputs or prior self-citations. The method is explicitly data-driven with training on undersampled cardiac cine scans and quantitative comparisons to baselines, rendering the central claims experimentally grounded rather than tautological.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that spatio-temporal redundancies in dynamic MRI are sufficient to recover signals from highly undersampled data via learned alternation between domains; no free parameters or invented entities are explicitly introduced beyond standard DL weights.

axioms (1)
  • domain assumption Dynamic MR imaging exhibits high correlations in k-space and time that can be exploited for reconstruction from undersampled data.
    Stated in the first sentence of the abstract as the basis for the approach.

pith-pipeline@v0.9.0 · 5716 in / 1196 out tokens · 29578 ms · 2026-05-24T17:40:10.213770+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.

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

18 extracted references · 18 canonical work pages · 1 internal anchor

  1. [1]

    Ak¸ cakaya, M., Moeller, S., Weing¨ artner, S., et al.: Scan-specific robust artificial- neural-networks for k-space interpolation (RAKI) reconstruction: Database-free deep learning for fast imaging. Magn. Reson. Med. 81(1), 439–453 (2019)

    work page 2019

  2. [2]

    Eo, T., Jun, Y., Kim, T., et al.: KIKI-net: cross-domain convolutional neural net- works for reconstructing undersampled magnetic resonance images. Magn. Reson. Med. 80(5), 2188–2201 (2018)

    work page 2018

  3. [3]

    Griswold, M.A., Jakob, P.M., Heidemann, R.M., et al.: Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magn. Reson. Med. 47(6), 1202–1210 (2002)

    work page 2002

  4. [4]

    Hammernik, K., Klatzer, T., Kobler, E., et al.: Learning a variational network for reconstruction of accelerated MRI data. Magn. Reson. Med. 79(6), 3055–3071 (2018)

    work page 2018

  5. [5]

    k-Space Deep Learning for Accelerated MRI

    Han, Y., Ye, J.C.: k-space deep learning for accelerated MRI. arXiv preprint arXiv:1805.03779 (2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  6. [6]

    Hu, X., Parrish, T.: Reduction of field of view for dynamic imaging. Magn. Reson. Med. 31(6), 691–694 (1994) k-t NEXT: k-t NEtwork with X-f Transform 9

    work page 1994

  7. [7]

    Jones, R., Haraldseth, O., M¨ uller, T., et al.: K-space substitution: a novel dynamic imaging technique. Magn. Reson. Med. 29(6), 830–834 (1993)

    work page 1993

  8. [8]

    Jung, H., Sung, K., Nayak, K.S., et al.: k-t FOCUSS: a general compressed sensing framework for high resolution dynamic MRI. Magn. Reson. Med. 61(1), 103–116 (2009)

    work page 2009

  9. [9]

    Physics in Medicine & Biology 52(11), 3201 (2007)

    Jung, H., Ye, J.C., Kim, E.Y.: Improved k–t BLAST and k–t SENSE using FO- CUSS. Physics in Medicine & Biology 52(11), 3201 (2007)

    work page 2007

  10. [10]

    IEEE Trans

    Lingala, S.G., Hu, Y., DiBella, E., et al.: Accelerated dynamic MRI exploiting sparsity and low-rank structure: k-t SLR. IEEE Trans. Med. Imaging 30(5), 1042– 1054 (2011)

    work page 2011

  11. [11]

    In: MICCAI

    Liu, J., Kuang, T., Zhang, X.: Image reconstruction by splitting deep learning regularization from iterative inversion. In: MICCAI. pp. 224–231. Springer (2018)

    work page 2018

  12. [12]

    IEEE Trans

    Qin, C., Schlemper, J., Caballero, J., et al.: Convolutional recurrent neural net- works for dynamic MR image reconstruction. IEEE Trans. Med. Imaging 38(1), 280–290 (2019)

    work page 2019

  13. [13]

    IEEE Trans

    Schlemper, J., Caballero, J., Hajnal, J.V., et al.: A deep cascade of convolutional neural networks for dynamic MR image reconstruction. IEEE Trans. Med. Imaging 37(2), 491–503 (2018)

    work page 2018

  14. [14]

    In: MLMIR

    Schlemper, J., Castro, D.C., Bai, W., et al.: Bayesian deep learning for accelerated MR image reconstruction. In: MLMIR. pp. 64–71. Springer (2018)

    work page 2018

  15. [15]

    Tsao, J., Boesiger, P., Pruessmann, K.P.: k-t BLAST and k-t SENSE: dynamic MRI with high frame rate exploiting spatiotemporal correlations. Magn. Reson. Med. 50(5), 1031–1042 (2003)

    work page 2003

  16. [16]

    Ye, J.C., Han, Y., Cha, E.: Deep convolutional framelets: A general deep learning framework for inverse problems. SIAM J. Imaging Sci. 11(2), 991–1048 (2018)

    work page 2018

  17. [17]

    In: MICCAI

    Zhang, P., Wang, F., Xu, W., et al.: Multi-channel generative adversarial network for parallel magnetic resonance image reconstruction in k-space. In: MICCAI. pp. 180–188 (2018)

    work page 2018

  18. [18]

    Zhang, S., Block, K.T., Frahm, J.: Magnetic resonance imaging in real time: ad- vances using radial FLASH. Magn. Reson. Med. 31(1), 101–109 (2010)

    work page 2010