pith. sign in

arxiv: 2606.19182 · v1 · pith:K2ZO6NOPnew · submitted 2026-06-17 · 📡 eess.IV

Optimized Multi-Contrast Self-Supervised MRI Reconstruction using Learned k-space Partitioning

Brenden Kadota , Charles Millard , Mark Chiew This is my paper

Pith reviewed 2026-06-26 18:59 UTC · model grok-4.3

classification 📡 eess.IV
keywords multi-contrast MRIself-supervised learningk-space partitioningMRI reconstructiondeep learningaccelerated MRIundersampled k-space
0
0 comments X

The pith

Multi-contrast self-supervised MRI reconstruction with learned k-space partitioning produces higher quality images than single-contrast methods without fully sampled reference data.

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

The paper introduces a training framework that uses under-sampled data from several MRI contrasts at once to reconstruct images. It replaces fixed splits of k-space with a learned probability distribution that generates the training masks in an end-to-end fashion. This removes the usual requirement for fully sampled scans during training. Tests on two public multi-contrast datasets show measurable gains in reconstruction accuracy over existing single-contrast self-supervised approaches. Learning the partition itself adds a further improvement to image fidelity.

Core claim

Jointly training a reconstruction network on multiple under-sampled contrasts while learning an optimal partitioning probability distribution, which is sampled to produce masks, improves reconstruction fidelity compared with single-contrast self-supervised learning and does so without any fully sampled k-space reference.

What carries the argument

A learned partitioning probability distribution that is sampled to generate masks dividing k-space data for self-supervised training across multiple contrasts.

If this is right

  • Reconstruction quality increases when multiple contrasts are used together under the learned-partition regime.
  • Training becomes possible on clinical multi-contrast scans that lack fully sampled references.
  • Protocol times can shorten because higher acceleration factors become usable while maintaining fidelity.
  • The same network can be applied to new contrasts once the partitioning distribution has been learned.

Where Pith is reading between the lines

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

  • The learned distribution may differ across contrasts, implying that optimal self-supervised splits are contrast-dependent rather than universal.
  • The framework could be tested on single-contrast data to isolate whether the learned partitioning alone improves results over fixed splits.
  • Larger clinical archives of multi-contrast exams could now serve as training sources even when full sampling was never performed.

Load-bearing premise

Joint training on multiple under-sampled contrasts with a learned partitioning distribution produces stable and unbiased reconstructions without a fully sampled reference.

What would settle it

Running the proposed method and standard single-contrast SSDU on the same two public multi-contrast datasets and finding no difference in SSIM or NRMSE on held-out test slices would falsify the improvement claim.

read the original abstract

Objective: Deep Learning has shown promise in accelerating MRI by reconstructing high-quality images from under-sampled data. While recent work has leveraged multi-contrast information to improve reconstruction performance, these methods rely on supervised learning, which requires fully sampled k-space for training. One method, self-supervised learning via data undersampling (SSDU), enables direct training on under-sampled k-space by partitioning it into two sets, with a network mapping between the two. In this work, we improve MRI self-supervised MRI reconstruction with two modifications. Methods: We propose a multi-contrast self-supervised learning framework that jointly trains on multiple under-sampled contrasts without requiring fully sampled k-space data as a reference. Moreover, we learn an optimal self-supervised data partitioning for each contrast in an end-to-end manner, further enhancing reconstruction quality. Specifically, we learn an optimal partitioning probability distribution, which is sampled to generate a mask for partitioning. Results: Experiments on two publicly available multi-contrast MRI datasets demonstrate the improved reconstruction quality of our proposed self-supervised multi-contrast learned partitioning method compared to the current single-contrast self-supervised learning methods. We also demonstrate that learning the partitioning of k-space data further enhances the fidelity of reconstructions. Conclusion: Multi-contrast reconstruction combined with learned partitioning improves reconstruction fidelity over single-contrast self-supervised MRI reconstructions. Significance: Our method can facilitate higher image fidelity and/or accelerated MRI protocol times compared to previous self-supervised methods, and without requiring fully sampled k-space for training.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 2 minor

Summary. The manuscript proposes extending self-supervised learning via data undersampling (SSDU) to a multi-contrast setting by jointly training a reconstruction network across multiple under-sampled contrasts and learning an optimal per-contrast k-space partitioning probability distribution in an end-to-end manner; the learned distribution is sampled to create the two SSDU partitions. Experiments on two publicly available multi-contrast MRI datasets are reported to demonstrate improved reconstruction quality relative to single-contrast SSDU baselines, with the learned partitioning providing further gains. No fully sampled reference data are used for training or supervision.

Significance. If the empirical claims hold after proper validation, the work would provide a practical route to multi-contrast self-supervised MRI reconstruction that avoids the need for fully sampled training data while optimizing the data-partitioning step, potentially enabling higher acceleration factors or improved fidelity in clinical protocols. The end-to-end learned partitioning is a clear technical extension of prior SSDU methods.

major comments (3)
  1. [Abstract/Results] Abstract, Results paragraph: the central claim of 'improved reconstruction quality' and 'enhanced fidelity' is stated without any quantitative metrics, PSNR/SSIM values, error bars, statistical tests, or ablation tables, preventing assessment of whether the reported gains are load-bearing or statistically meaningful.
  2. [Methods] Methods paragraph (learned partitioning): the loss compares only the two learned partitions with no full k-space term; the manuscript provides no analysis or hold-out fully sampled validation set demonstrating that systematic biases consistent across the learned masks are not absorbed into the network weights, leaving the unbiased-reconstruction assumption unverified.
  3. [Methods/Results] Methods/Results: no description of convergence behavior of the learned partitioning distribution, stability across random seeds, or comparison against fixed (non-learned) multi-contrast SSDU masks, so it is unclear whether the end-to-end optimization yields a stable fixed point rather than a data-dependent artifact.
minor comments (2)
  1. [Abstract] Abstract: the phrase 'learned k-space Partitioning' is capitalized inconsistently with the rest of the text.
  2. [Abstract] Abstract: the significance statement repeats the results claim without adding new information about clinical impact or limitations.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for the constructive comments on our manuscript. We respond point-by-point to the major comments below, committing to revisions where they strengthen the work without altering its core claims.

read point-by-point responses

  1. Referee: [Abstract/Results] Abstract, Results paragraph: the central claim of 'improved reconstruction quality' and 'enhanced fidelity' is stated without any quantitative metrics, PSNR/SSIM values, error bars, statistical tests, or ablation tables, preventing assessment of whether the reported gains are load-bearing or statistically meaningful.

    Authors: We agree the abstract would be strengthened by quantitative support. The results section contains PSNR/SSIM tables and comparisons on both datasets; we will revise the abstract to report key metric improvements with error bars and will add explicit references to statistical tests and ablation tables in the results. revision: yes

  2. Referee: [Methods] Methods paragraph (learned partitioning): the loss compares only the two learned partitions with no full k-space term; the manuscript provides no analysis or hold-out fully sampled validation set demonstrating that systematic biases consistent across the learned masks are not absorbed into the network weights, leaving the unbiased-reconstruction assumption unverified.

    Authors: Our loss follows the SSDU formulation, which by design uses only the two partitions because fully sampled k-space is unavailable. We cannot supply a hold-out fully sampled validation set, as none exists in the experimental protocol. We will expand the discussion to explicitly address the unbiased-reconstruction assumption and its limitations in the self-supervised regime. revision: partial

  3. Referee: [Methods/Results] Methods/Results: no description of convergence behavior of the learned partitioning distribution, stability across random seeds, or comparison against fixed (non-learned) multi-contrast SSDU masks, so it is unclear whether the end-to-end optimization yields a stable fixed point rather than a data-dependent artifact.

    Authors: We will add the requested analyses in revision: training curves for the partitioning distribution, results across multiple random seeds, and direct comparisons against fixed (non-learned) multi-contrast SSDU masks to demonstrate stability of the learned solution. revision: yes

standing simulated objections not resolved
  • Verification of the unbiased-reconstruction assumption via hold-out fully sampled data, which is unavailable by design of the self-supervised method.

Circularity Check

0 steps flagged

No significant circularity; empirical method with independent experimental validation

full rationale

The paper extends SSDU-style self-supervised training to multi-contrast data by jointly optimizing a network across contrasts while learning per-contrast partitioning distributions end-to-end. The central claims consist of measured reconstruction quality improvements on two public datasets, obtained by comparing the proposed method against single-contrast baselines. No equations or steps in the provided description reduce the reported gains to a fitted parameter, a self-referential loss term, or a self-citation chain; the training loss operates between data partitions while evaluation metrics are computed on held-out data. The derivation chain therefore remains self-contained against external benchmarks rather than tautological.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides insufficient technical detail to identify specific free parameters, axioms, or invented entities beyond the general description of a learned partitioning probability distribution.

pith-pipeline@v0.9.1-grok · 5793 in / 1078 out tokens · 28980 ms · 2026-06-26T18:59:24.550189+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

56 extracted references · 44 canonical work pages

  1. [1]

    Griswold, Peter M

    M. A. Griswold et al. , “Generalized autocalibrating partially parallel acquisitions (GRAPPA),” Magn. Reson. Med., vol. 47, no. 6, pp. 1202 –1210, 2002, doi: 10.1002/mrm.10171

    work page doi:10.1002/mrm.10171 2002

  2. [2]

    SENSE: Sensitivity encoding for fast MRI,

    K. P. Pruessmann et al., “SENSE: Sensitivity encoding for fast MRI,” Magn. Reson. Med. , vol. 42, no. 5, pp. 952–962, 1999, doi: 10.1002/(SICI)1522 - 2594(199911)42:5<952::AID-MRM16>3.0.CO;2-S

    work page doi:10.1002/(sici)1522 1999

  3. [3]

    Compressed Sensing MRI,

    M. Lustig et al. , “Compressed Sensing MRI,” IEEE Signal Process. Mag. , vol. 25, no. 2, pp. 72 –82, Mar. 2008, doi: 10.1109/MSP.2007.914728

    work page doi:10.1109/msp.2007.914728 2008

  4. [4]

    Improving parallel imaging by jointly reconstructing multi-contrast data,

    B. Bilgic et al., “Improving parallel imaging by jointly reconstructing multi-contrast data,” Magn. Reson. Med., vol. 80, no. 2, pp. 619 –632, 2018, doi: 10.1002/mrm.27076

    work page doi:10.1002/mrm.27076 2018

  5. [5]

    Deep -Learning-Based Multi -Modal Fusion for Fast MR Reconstruction,

    L. Xiang et al. , “Deep -Learning-Based Multi -Modal Fusion for Fast MR Reconstruction,” IEEE Trans. Biomed. Eng., vol. 66, no. 7, pp. 2105 –2114, Jul. 2019, doi: 10.1109/TBME.2018.2883958

    work page doi:10.1109/tbme.2018.2883958 2019

  6. [6]

    Prior-Guided Image Reconstruction for Accelerated Multi -Contrast MRI via Generative Adversarial Networks,

    S. U. H. Dar et al., “Prior-Guided Image Reconstruction for Accelerated Multi -Contrast MRI via Generative Adversarial Networks,” IEEE J. Sel. Top. Signal Process., vol. 14, no. 6, pp. 1072 –1087, Oct. 2020, doi: 10.1109/JSTSP.2020.3001737

    work page doi:10.1109/jstsp.2020.3001737 2020

  7. [7]

    Deep unregistered multi -contrast MRI reconstruction,

    X. Liu et al. , “Deep unregistered multi -contrast MRI reconstruction,” Magn. Reson. Imaging, vol. 81, pp. 33– 41, Sep. 2021, doi: 10.1016/j.mri.2021.05.005

    work page doi:10.1016/j.mri.2021.05.005 2021

  8. [8]

    DuDoCAF: Dual-Domain Cross-Attention Fusion with Recurrent Transformer for Fast Multi - contrast MR Imaging,

    J. Lyu et al., “DuDoCAF: Dual-Domain Cross-Attention Fusion with Recurrent Transformer for Fast Multi - contrast MR Imaging,” in Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 , L. Wang, Q. Dou, P. T. Fletcher, S. Speidel, and S. Li, Eds., Cham: Springer Nature Switzerland, 2022, pp. 474 –484. doi: 10.1007/978-3-031-16446-0_45

    work page doi:10.1007/978-3-031-16446-0_45 2022

  9. [9]

    Ultra -Fast T2 -Weighted MR Reconstruction Using Complementary T1 -Weighted Information,

    L. Xiang et al. , “Ultra -Fast T2 -Weighted MR Reconstruction Using Complementary T1 -Weighted Information,” Med. Image Comput. Comput. -Assist. Interv. MICCAI Int. Conf. Med. Image Comput. Comput.- Assist. Interv., vol. 11070, pp. 215 –223, Sep. 2018, doi: 10.1007/978-3-030-00928-1_25

    work page doi:10.1007/978-3-030-00928-1_25 2018

  10. [10]

    Multi -contrast reconstruction with Bayesian compressed sensing,

    B. Bilgic et al. , “Multi -contrast reconstruction with Bayesian compressed sensing,” Magn. Reson. Med., vol. 66, no. 6, pp. 1601 –1615, 2011, doi: 10.1002/mrm.22956

    work page doi:10.1002/mrm.22956 2011

  11. [11]

    Vectorial total generalized variation for accelerated multi -channel multi -contrast MRI,

    I. Chatnuntawech et al. , “Vectorial total generalized variation for accelerated multi -channel multi -contrast MRI,” Magn. Reson. Imaging , vol. 34, no. 8, pp. 1161 – 1170, 2016, doi: 10.1016/j.mri.2016.05.014

    work page doi:10.1016/j.mri.2016.05.014 2016

  12. [12]

    Sparse MRI reconstruction using multi - contrast image guided graph representation,

    Z. Lai et al. , “Sparse MRI reconstruction using multi - contrast image guided graph representation,” Magn. Reson. Imaging , vol. 43, pp. 95 –104, Nov. 2017, doi: 10.1016/j.mri.2017.07.009

    work page doi:10.1016/j.mri.2017.07.009 2017

  13. [13]

    High -dimensionality undersampled patch-based reconstruction (HD-PROST) for accelerated multi-contrast MRI,

    A. Bustin et al. , “High -dimensionality undersampled patch-based reconstruction (HD-PROST) for accelerated multi-contrast MRI,” Magn. Reson. Med., vol. 81, no. 6, pp. 3705–3719, 2019, doi: 10.1002/mrm.27694

    work page doi:10.1002/mrm.27694 2019

  14. [14]

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

    X. Qu et al., “Magnetic resonance image reconstruction from undersampled measurements using a patch -based nonlocal operator,” Med. Image Anal., vol. 18, no. 6, pp. 843–856, Aug. 2014, doi: 10.1016/j.media.2013.09.007

    work page doi:10.1016/j.media.2013.09.007 2014

  15. [15]

    End-to-End Variational Networks for Accelerated MRI Reconstruction,

    A. Sriram et al., “End-to-End Variational Networks for Accelerated MRI Reconstruction,” Apr. 15, 2020, arXiv: arXiv:2004.06688. Accessed: Jan. 09, 2023. [Online]. Available: http://arxiv.org/abs/2004.06688

    arXiv 2020

  16. [16]

    Aggarwal, Merry P

    H. K. Aggarwal et al. , “MoDL: Model Based Deep Learning Architecture for Inverse Problems,” IEEE Trans. Med. Imaging , vol. 38, no. 2, pp. 394 –405, Feb. 2019, doi: 10.1109/TMI.2018.2865356

    work page doi:10.1109/tmi.2018.2865356 2019

  17. [17]

    A Deep Information Sharing Network for Multi-Contrast Compressed Sensing MRI Reconstruction,

    L. Sun et al., “A Deep Information Sharing Network for Multi-Contrast Compressed Sensing MRI Reconstruction,” IEEE Trans. Image Process. , vol. 28, no. 12, pp. 6141 –6153, Dec. 2019, doi: 10.1109/TIP.2019.2925288

    work page doi:10.1109/tip.2019.2925288 2019

  18. [18]

    Joint multi-contrast variational network reconstruction (jVN) with application to rapid 2D and 3D imaging,

    D. Polak et al., “Joint multi-contrast variational network reconstruction (jVN) with application to rapid 2D and 3D imaging,” Magn. Reson. Med., vol. 84, no. 3, pp. 1456 – 1469, 2020, doi: 10.1002/mrm.28219

    work page doi:10.1002/mrm.28219 2020

  19. [19]

    Joint MAPLE: Accel- erated joint T1 and T2* mapping with scan-specific self-supervised networks.Magnetic Resonance in Medicine, 91(6):2294–2309, 2024

    A. Heydari et al., “Joint MAPLE: Accelerated joint T and mapping with scan -specific self -supervised networks,” Magn. Reson. Med., vol. 91, no. 6, pp. 2294–2309, 2024, doi: 10.1002/mrm.29989

    work page doi:10.1002/mrm.29989 2024

  20. [20]

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale,

    A. Dosovitskiy et al., “An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale,” Jun. 03, 2021, arXiv: arXiv:2010.11929. Accessed: Jan. 04, 2023. [Online]. Available: http://arxiv.org/abs/2010.11929

    Pith/arXiv arXiv 2021

  21. [21]

    Multimodal Transformer for Accelerated MR Imaging,

    C.-M. Feng et al. , “Multimodal Transformer for Accelerated MR Imaging,” IEEE Trans. Med. Imaging , vol. 42, no. 10, pp. 2804 –2816, Oct. 2023, doi: 10.1109/TMI.2022.3180228

    work page doi:10.1109/tmi.2022.3180228 2023

  22. [22]

    Dataset condensation with distribution matching

    B. Zhou et al. , “DSFormer: A Dual -domain Self - supervised Transformer for Accelerated Multi -contrast MRI Reconstruction,” in 2023 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), Waikoloa, HI, USA: IEEE, Jan. 2023, pp. 4955–4964. doi: 10.1109/WACV56688.2023.00494

    work page doi:10.1109/wacv56688.2023.00494 2023

  23. [23]

    Adapted random sampling patterns for accelerated MRI,

    F. Knoll et al., “Adapted random sampling patterns for accelerated MRI,” Magn. Reson. Mater. Phys. Biol. Med., vol. 24, no. 1, pp. 43 –50, Feb. 2011, doi: 10.1007/s10334-010-0234-7

    work page doi:10.1007/s10334-010-0234-7 2011

  24. [24]

    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,” Magn. Reson. Imaging, vol. 33, no. 5, pp. 635– 643, Jun. 2015, doi: 10.1016/j.mri.2015.01.008. > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE -CLICK HERE TO EDIT) < 12

    work page doi:10.1016/j.mri.2015.01.008 2015

  25. [25]

    Learning -Based Compressive MRI,

    B. Gözcü et al. , “Learning -Based Compressive MRI,” May 03, 2018, arXiv: arXiv:1805.01266. Accessed: Jan. 15, 2024. [Online]. Available: http://arxiv.org/abs/1805.01266

    Pith/arXiv arXiv 2018

  26. [26]

    Optimization of k-space trajectories for compressed sensing by Bayesian experimental design,

    M. Seeger et al., “Optimization of k-space trajectories for compressed sensing by Bayesian experimental design,” Magn. Reson. Med. , vol. 63, no. 1, pp. 116 –126, 2010, doi: 10.1002/mrm.22180

    work page doi:10.1002/mrm.22180 2010

  27. [27]

    Statistically Segregated k -Space Sampling for Accelerating Multiple -Acquisition MRI,

    L. K. Senel et al. , “Statistically Segregated k -Space Sampling for Accelerating Multiple -Acquisition MRI,” IEEE Trans. Med. Imaging , vol. 38, no. 7, pp. 1701 – 1714, Jul. 2019, doi: 10.1109/TMI.2019.2892378

    work page doi:10.1109/tmi.2019.2892378 2019

  28. [28]

    Fast Multi-Contrast MRI Acquisition by Optimal Sampling of Information Complementary to Pre-Acquired MRI Contrast,

    J. Yang et al., “Fast Multi-Contrast MRI Acquisition by Optimal Sampling of Information Complementary to Pre-Acquired MRI Contrast,” IEEE Trans. Med. Imaging, vol. 42, no. 5, pp. 1363 –1373, May 2023, doi: 10.1109/TMI.2022.3227262

    work page doi:10.1109/tmi.2022.3227262 2023

  29. [29]

    AutoSamp: Autoencoding k -space Sampling via Variational Information Maximization for 3D MRI,

    C. Alkan et al. , “AutoSamp: Autoencoding k -space Sampling via Variational Information Maximization for 3D MRI,” IEEE Trans. Med. Imaging, pp. 1–1, 2024, doi: 10.1109/TMI.2024.3443292

    work page doi:10.1109/tmi.2024.3443292 2024

  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 J. Sel. Top. Signal Process. , vol. 14, no. 6, pp. 1151 –1162, Oct. 2020, doi: 10.1109/JSTSP.2020.3004094

    work page doi:10.1109/jstsp.2020.3004094 2020

  31. [31]

    Learning Optimal K -Space Acquisition and Reconstruction Using Physics -Informed Neural Networks

    W. Peng et al., “Learning Optimal K -Space Acquisition and Reconstruction Using Physics -Informed Neural Networks”

  32. [32]

    B -Spline Parameterized Joint Optimization of Reconstruction and K -Space Trajectories (BJORK) for Accelerated 2D MRI,

    G. Wang et al. , “B -Spline Parameterized Joint Optimization of Reconstruction and K -Space Trajectories (BJORK) for Accelerated 2D MRI,” IEEE Trans. Med. Imaging, vol. 41, no. 9, pp. 2318–2330, Sep. 2022, doi: 10.1109/TMI.2022.3161875

    work page doi:10.1109/tmi.2022.3161875 2022

  33. [33]

    JoJoNet: Joint -contrast and Joint - sampling-and-reconstruction Network for Multi-contrast MRI,

    L. Zhao et al. , “JoJoNet: Joint -contrast and Joint - sampling-and-reconstruction Network for Multi-contrast MRI,” Oct. 26, 2022, arXiv: arXiv:2210.12548. Accessed: Feb. 17, 2023. [Online]. Available: http://arxiv.org/abs/2210.12548

    arXiv 2022

  34. [34]

    Wang, Adrian V

    C. D. Bahadir et al. , “Deep -Learning-Based Optimization of the Under -Sampling Pattern in MRI,” IEEE Trans. Comput. Imaging , vol. 6, pp. 1139 –1152, 2020, doi: 10.1109/TCI.2020.3006727

    work page doi:10.1109/tci.2020.3006727 2020

  35. [35]

    Extending LOUPE for K -Space Under- Sampling Pattern Optimization in Multi -coil MRI,

    J. Zhang et al., “Extending LOUPE for K -Space Under- Sampling Pattern Optimization in Multi -coil MRI,” in Machine Learning for Medical Image Reconstruction, F. Deeba, P. Johnson, T. Würfl, and J. C. Ye, Eds., Cham: Springer International Publishing, 2020, pp. 91–101. doi: 10.1007/978-3-030-61598-7_9

    work page doi:10.1007/978-3-030-61598-7_9 2020

  36. [36]

    Dual -domain accelerated MRI reconstruction using transformers with learning -based undersampling,

    G. Q. Hong et al. , “Dual -domain accelerated MRI reconstruction using transformers with learning -based undersampling,” Comput. Med. Imaging Graph. , vol. 106, p. 102206, Jun. 2023, doi: 10.1016/j.compmedimag.2023.102206

    work page doi:10.1016/j.compmedimag.2023.102206 2023

  37. [37]

    Auto -Encoding Variational Bayes,

    D. P. Kingma and M. Welling, “Auto -Encoding Variational Bayes,” 2017. Accessed: Nov. 11, 2024. [Online]. Available: http://arxiv.org/abs/1312.6114

    Pith/arXiv arXiv 2017

  38. [38]

    Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation,

    Y. Bengio et al., “Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation,” Aug. 15, 2013, arXiv: arXiv:1308.3432. Accessed: Nov. 04, 2024. [Online]. Available: http://arxiv.org/abs/1308.3432

    Pith/arXiv arXiv 2013

  39. [39]

    Self-supervised learning of physics-guided reconstruction neural networks without fully sampled reference data.Magnetic Resonance in Medicine, 84 (6):3172–3191, 2020

    B. Yaman et al. , “Self -supervised learning of physics - guided reconstruction neural networks without fully sampled reference data,” Magn. Reson. Med., vol. 84, no. 6, pp. 3172–3191, 2020, doi: 10.1002/mrm.28378

    work page doi:10.1002/mrm.28378 2020

  40. [40]

    Multi -mask self -supervised learning for physics-guided neural networks in highly accelerated magnetic resonance imaging,

    B. Yaman et al. , “Multi -mask self -supervised learning for physics-guided neural networks in highly accelerated magnetic resonance imaging,” NMR Biomed., vol. 35, no. 12, p. e4798, 2022, doi: 10.1002/nbm.4798

    work page doi:10.1002/nbm.4798 2022

  41. [41]

    Duncan, and Michal Sofka

    B. Zhou et al., “Dual-domain self-supervised learning for accelerated non -Cartesian MRI reconstruction,” Med. Image Anal. , vol. 81, p. 102538, Oct. 2022, doi: 10.1016/j.media.2022.102538

    work page doi:10.1016/j.media.2022.102538 2022

  42. [42]

    A Theoretical Framework for Self-Supervised MR Image Reconstruction Using Sub - Sampling via Variable Density Noisier2Noise,

    C. Millard and M. Chiew, “A Theoretical Framework for Self-Supervised MR Image Reconstruction Using Sub - Sampling via Variable Density Noisier2Noise,” IEEE Trans. Comput. Imaging, vol. 9, pp. 707–720, 2023, doi: 10.1109/TCI.2023.3299212

    work page doi:10.1109/tci.2023.3299212 2023

  43. [43]

    U-Net: Convolutional Networks for Biomedical Image Segmentation,

    O. Ronneberger et al., “U-Net: Convolutional Networks for Biomedical Image Segmentation,” May 18, 2015, arXiv: arXiv:1505.04597. Accessed: Feb. 28, 2023. [Online]. Available: http://arxiv.org/abs/1505.04597

    Pith/arXiv arXiv 2015

  44. [44]

    The RSNA-ASNR-MICCAI BraTS 2021 Benchmark on Brain Tumor Segmentation and Radiogenomic Classification,

    U. Baid et al., “The RSNA-ASNR-MICCAI BraTS 2021 Benchmark on Brain Tumor Segmentation and Radiogenomic Classification,” Sep. 12, 2021, arXiv: arXiv:2107.02314. Accessed: Feb. 28, 2023. [Online]. Available: http://arxiv.org/abs/2107.02314

    Pith/arXiv arXiv 2021

  45. [45]

    Coil compression for accelerated imaging with Cartesian sampling,

    T. Zhang et al. , “Coil compression for accelerated imaging with Cartesian sampling,” Magn. Reson. Med. , vol. 69, no. 2, pp. 571 –582, 2013, doi: 10.1002/mrm.24267

    work page doi:10.1002/mrm.24267 2013

  46. [46]

    M4Raw: A multi -contrast, multi - repetition, multi -channel MRI k -space dataset for low - field MRI research,

    M. Lyu et al. , “M4Raw: A multi -contrast, multi - repetition, multi -channel MRI k -space dataset for low - field MRI research,” Sci. Data, vol. 10, no. 1, Art. no. 1, May 2023, doi: 10.1038/s41597-023-02181-4

    work page doi:10.1038/s41597-023-02181-4 2023

  47. [47]

    Adam: A Method for Stochastic Optimization,

    D. P. Kingma and J. Ba, “Adam: A Method for Stochastic Optimization,” Jan. 30, 2017, arXiv: arXiv:1412.6980. Accessed: Nov. 05, 2024. [Online]. Available: http://arxiv.org/abs/1412.6980

    Pith/arXiv arXiv 2017

  48. [48]

    fastMRI: An Open Dataset and Benchmarks for Accelerated MRI,

    J. Zbontar et al. , “fastMRI: An Open Dataset and Benchmarks for Accelerated MRI,” Dec. 11, 2019, arXiv: arXiv:1811.08839. Accessed: Feb. 28, 2023. [Online]. Available: http://arxiv.org/abs/1811.08839

    Pith/arXiv arXiv 2019

  49. [49]

    A multi -layer transcranial focused ultrasound model for neuromodulation procedure planning and insertion loss estimation

    Y. Yan et al. , “IWNeXt: an image -wavelet domain ConvNeXt-based network for self -supervised multi - contrast MRI reconstruction,” Phys. Med. Biol., vol. 69, no. 8, p. 085005, Apr. 2024, doi: 10.1088/1361 - 6560/ad33b4

    work page doi:10.1088/1361 2024

  50. [50]

    Multi -Modal MRI Reconstruction Assisted with Spatial Alignment Network,

    K. Xuan et al. , “Multi -Modal MRI Reconstruction Assisted with Spatial Alignment Network,” IEEE Trans. Med. Imaging, vol. 41, no. 9, pp. 2499–2509, Sep. 2022, doi: 10.1109/TMI.2022.3164050

    work page doi:10.1109/tmi.2022.3164050 2022

  51. [51]

    Robots and AI: Illusions and Social Dilemmas

    C. Hu et al. , “Self -supervised Learning for MRI Reconstruction with a Parallel Network Training Framework,” in Medical Image Computing and Computer Assisted Intervention – MICCAI 2021, M. de Bruijne, P. C. Cattin, S. Cotin, N. Padoy, S. Speidel, Y. Zheng, and C. Essert, Eds., Cham: Springer International > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATIO...

    work page doi:10.1007/978-3-030- 2021

  52. [52]

    Clean Self -Supervised MRI Reconstruction from Noisy, Sub-Sampled Training Data with Robust SSDU,

    C. Millard and M. Chiew, “Clean Self -Supervised MRI Reconstruction from Noisy, Sub-Sampled Training Data with Robust SSDU,” Bioengineering, vol. 11, no. 12, p. 1305, Dec. 2024, doi: 10.3390/bioengineering11121305

    work page doi:10.3390/bioengineering11121305 2024

  53. [53]

    Noise2Recon: Enabling SNR -robust MRI reconstruction with semi -supervised and self - supervised learning,

    A. D. Desai et al., “Noise2Recon: Enabling SNR -robust MRI reconstruction with semi -supervised and self - supervised learning,” Magn. Reson. Med., vol. 90, no. 5, pp. 2052–2070, 2023, doi: 10.1002/mrm.29759

    work page doi:10.1002/mrm.29759 2052

  54. [54]

    Noise2Contrast: Multi -Contrast Fusion Enables Self -Supervised Tomographic Image Denoising,

    F. Wagner et al. , “Noise2Contrast: Multi -Contrast Fusion Enables Self -Supervised Tomographic Image Denoising,” Dec. 09, 2022, arXiv: arXiv:2212.04832. Accessed: Feb. 28, 2023. [Online]. Available: http://arxiv.org/abs/2212.04832

    arXiv 2022

  55. [55]

    Reconstruction of multicontrast MR images through deep learning,

    W.-J. Do et al. , “Reconstruction of multicontrast MR images through deep learning,” Med. Phys., vol. 47, no. 3, pp. 983–997, 2020, doi: 10.1002/mp.14006

    work page doi:10.1002/mp.14006 2020

  56. [56]

    Validation and Generalizability of Self - Supervised Image Reconstruction Methods for Undersampled MRI,

    T. Yu et al. , “Validation and Generalizability of Self - Supervised Image Reconstruction Methods for Undersampled MRI,” Mach. Learn. Biomed. Imaging , vol. 1, no. September 2022, pp. 1 –31, Sep. 2022, doi: 10.59275/j.melba.2022-6g33

    work page doi:10.59275/j.melba.2022-6g33 2022