arxiv: 2604.01710 · v2 · submitted 2026-04-02 · ⚛️ physics.med-ph

Recognition: no theorem link

High-resolution ultra-low-field MRI with SNRAware denoising

Teresa Guallart-Naval , Hui Xue , Jos\'e M. Algar\'in , Eli G. Castanon , Jes\'us Conejero , Fernando Galve , Mary A. Nassejje , John Stairs

show 3 more authors

Lorena Vega-Cid Michael Hansen Joseba Alonso

Authors on Pith no claims yet

Pith reviewed 2026-05-13 21:04 UTC · model grok-4.3

classification ⚛️ physics.med-ph

keywords ultra-low-field MRIdeep learning denoisingSNRAwaresignal-to-noise ratioimage enhancementportable MRIhigh-resolution imaging

0 comments

The pith

Deep learning denoising with SNRAware raises effective SNR in ultra-low-field MRI to enable clinical-resolution images.

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

The paper evaluates a deep learning denoising model trained via the SNRAware framework on ultra-low-field MRI data at 72 mT and 88 mT. It shows that the model reliably boosts signal-to-noise ratio across various acquisition parameters, allowing nominal spatial resolutions that match those in standard 3 T clinical scans. Residual analysis confirms the model removes mainly random noise while keeping anatomical signal intact. Performance varies with initial SNR and can be affected by mismatches between training and test data. This approach points toward making low-cost, portable ULF systems more viable for broader use.

Core claim

The central claim is that the SNRAware-trained DL model consistently increases the effective SNR of ULF acquisitions, enabling images with nominal spatial resolutions comparable to those commonly used in clinical 3 T protocols, while predominantly removing stochastic noise and preserving signal structure.

What carries the argument

The SNRAware framework, a training method for deep learning denoising models that accounts for signal-to-noise ratio awareness.

Load-bearing premise

The assumption that the noise removal observed in residuals will translate to preserved diagnostic information across all real-world scanning conditions and patient anatomies.

What would settle it

A controlled study comparing diagnostic accuracy of radiologists reading denoised ULF images versus standard high-field images for a specific pathology like brain lesions.

Figures

Figures reproduced from arXiv: 2604.01710 by Eli G. Castanon, Fernando Galve, Hui Xue, Jes\'us Conejero, John Stairs, Joseba Alonso, Jos\'e M. Algar\'in, Lorena Vega-Cid, Mary A. Nassejje, Michael Hansen, Teresa Guallart-Naval.

**Figure 4.** Figure 4: Representative results from Experiment 2. Sagittal knee images at [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

**Figure 5.** Figure 5: Representative results from Experiments 3–6. ACR phantom reconstructions comparing raw FFT and SNRAware under varying acquisition conditions. (a) [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗

**Figure 6.** Figure 6: Representative results from Experiment 7. Raw FFT magnitude reconstructions and corresponding SNRAware-denoised images acquired under different [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗

**Figure 7.** Figure 7: Representative results from Experiment 8. Raw FFT magnitude reconstructions (left column) and the corresponding BM4D- and SNRAware-denoised [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗

**Figure 8.** Figure 8: Representative results from Experiment 9. ACR phantom images acquired on two ULF-MRI systems: NextMRI (88 mT) and Physio I (72 mT). Columns [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗

**Figure 9.** Figure 9: Representative results from Experiment 10. The raw images are shown in the top, with the corresponding denoised images below. This complete knee [PITH_FULL_IMAGE:figures/full_fig_p008_9.png] view at source ↗

read the original abstract

Ultra-low-field (ULF, <0.1 T) magnetic resonance imaging (MRI) systems offer advantages in cost, portability, and accessibility, but their current utility is still limited by low signal-to-noise ratio (SNR). Deep learning (DL)-based denoising has emerged as a potential strategy to mitigate this limitation. In this work, we present a systematic evaluation of a high-performance DL denoising model trained using the SNRAware framework and applied to 88 mT and 72 mT data. Using a series of controlled experiments, we assessed model performance as a function of spatial resolution, coil impedance matching, readout bandwidth, input noise level, k-space undersampling, anatomy, image contrast, and scanner platform, and compared against analytical denoising algorithms. The model consistently increased the effective SNR of ULF acquisitions, enabling images with nominal spatial resolutions comparable to those commonly used in clinical 3 T protocols. Residual analyses indicated that the model predominantly removed stochastic noise while preserving underlying signal structure. At the same time, the results highlight some constraints: denoising performance remains dependent on the starting SNR of the acquisition, and training-domain mismatch influences behavior under certain artifact conditions. These findings suggest that DL-based denoising can significantly expand the practical capabilities of ULF MRI, while emphasizing potential benefits from hardware-software co-optimization and the need for rigorous clinical validation to determine the diagnostic value of denoised images.

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.

Desk Editor's Note private letter to a colleague

The paper shows SNR-aware DL denoising can push 72-88 mT MRI to resolutions overlapping clinical 3 T protocols in controlled tests, but gains depend on input SNR and training match.

read the letter

This paper's main result is that a denoising model trained under the SNRAware framework lifts effective SNR on 72-88 mT scans enough to support nominal resolutions that line up with common 3 T clinical settings. The authors run a systematic set of controlled experiments that change resolution, coil matching, bandwidth, noise level, undersampling, anatomy, contrast, and platform, then compare the output to analytical denoising and inspect residuals. The residuals indicate the network mostly removes stochastic noise while keeping signal structure, which is a useful check. They also note the practical constraints: performance falls when starting SNR is low and can shift under artifact conditions that differ from the training set. These caveats are stated plainly rather than buried. The work is empirical and bounded, with no over-claim that the denoised images are diagnostically equivalent yet. The soft spots are the lack of reported quantitative metrics, error bars, or sample sizes in the summary, which makes the size of the gains hard to judge from the abstract alone. If the full text supplies those numbers and the data are available, the claim becomes easier to assess. The paper is aimed at groups building or using portable low-field MRI systems and at researchers working on DL denoising for medical images. A reader who needs to know how existing denoising methods behave under real ULF acquisition variables will find the multi-factor evaluation helpful. The experiments are systematic enough and the claims are sufficiently hedged that the paper deserves a serious referee rather than a desk rejection. I would send it to review with instructions to focus on the quantitative reporting and the clinical validation step.

Referee Report

1 major / 3 minor

Summary. The manuscript presents a systematic evaluation of a deep learning denoising model trained with the SNRAware framework and applied to 88 mT and 72 mT ULF MRI data. Controlled experiments assess performance as a function of spatial resolution, coil impedance matching, readout bandwidth, input noise level, k-space undersampling, anatomy, contrast, and scanner platform, with comparisons to analytical denoising methods. The central claim is that the model consistently increases effective SNR, enabling nominal spatial resolutions comparable to clinical 3 T protocols, supported by residual analyses showing predominant stochastic noise removal while preserving signal structure; performance dependencies on starting SNR and training-domain mismatch are explicitly noted, along with a call for clinical validation.

Significance. If the reported SNR gains and noise-removal behavior hold under broader conditions, this work could meaningfully expand the clinical reach of ULF MRI by addressing its primary SNR limitation through software means, thereby supporting the advantages of low cost, portability, and accessibility. The controlled, multi-factor experimental design and direct comparison to analytical baselines provide a reproducible basis for assessing the approach, and the explicit acknowledgment of constraints (SNR dependence, domain mismatch) strengthens the manuscript's credibility.

major comments (1)

Results section: the central claim that the model 'consistently increased the effective SNR' and enables 3 T-comparable nominal resolutions is load-bearing, yet the reported experiments lack specific quantitative metrics (e.g., measured SNR gain values, error bars, or statistical significance across the tested conditions), which are needed to evaluate effect sizes and reproducibility.

minor comments (3)

Abstract: the phrase 'effective SNR' is used without a concise definition or reference to the exact computation method employed in the residual analyses.
Discussion: the dependence on starting SNR and training-domain mismatch is acknowledged but could be illustrated with a dedicated figure or table summarizing performance degradation under specific artifact conditions.
Methods: clarify whether the SNRAware training framework introduces any additional hyperparameters beyond standard denoising networks and how they were selected.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive review and the recommendation for minor revision. We address the single major comment below.

read point-by-point responses

Referee: Results section: the central claim that the model 'consistently increased the effective SNR' and enables 3 T-comparable nominal resolutions is load-bearing, yet the reported experiments lack specific quantitative metrics (e.g., measured SNR gain values, error bars, or statistical significance across the tested conditions), which are needed to evaluate effect sizes and reproducibility.

Authors: We agree that tabulating explicit SNR gain values, including variability measures and statistical tests, would strengthen the presentation of the central claim. The current manuscript relies on visual comparisons, residual maps, and qualitative statements of improvement across conditions. In the revised manuscript we will add a new table in the Results section that reports measured effective SNR (pre- and post-denoising) for each tested parameter combination, together with standard deviations across repeated acquisitions and p-values from paired statistical tests. Error bars will also be added to the relevant summary plots. These additions will allow direct evaluation of effect sizes and reproducibility without altering the experimental design or conclusions. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper presents a systematic empirical evaluation of a DL-based denoising model applied to ULF MRI data, using controlled experiments across resolution, noise levels, undersampling, anatomy, contrast, and platforms, with direct comparisons to analytical denoising methods. Performance claims rest on measured SNR increases and residual analyses showing stochastic noise removal, without any mathematical derivations, predictions, or parameter fits that reduce to inputs by construction. Explicit caveats on SNR dependence and domain mismatch are reported, and the work calls for external clinical validation rather than asserting equivalence. This is a self-contained empirical study against external benchmarks with no load-bearing self-citations or self-definitional steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Central claim rests on the domain assumption that a DL model trained via SNRAware generalizes to remove stochastic noise in ULF acquisitions without introducing diagnostic artifacts; no free parameters or invented entities are introduced.

axioms (1)

domain assumption Deep learning models trained on SNRAware data can be applied to 72 mT and 88 mT ULF MRI acquisitions while predominantly removing stochastic noise.
Invoked in the description of model performance across acquisition parameters and residual analyses.

pith-pipeline@v0.9.0 · 5597 in / 1142 out tokens · 36926 ms · 2026-05-13T21:04:48.065100+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages

[1]

Aja-Fern ´andez and G

S. Aja-Fern ´andez and G. Vegas-S ´anchez-Ferrero,Statistical Analysis of Noise in MRI. Springer, 2016

work page 2016
[2]

Tackling SNR at low-field: a review of hardware approaches for point-of-care systems,

A. Webb and T. O’Reilly, “Tackling SNR at low-field: a review of hardware approaches for point-of-care systems,”Magnetic Resonance Materials in Physics, Biology and Medicine, vol. 36, no. 3, pp. 375–393, 2023

work page 2023
[3]

Design and operation of a whole-body MRI scanner without RF shielding,

S. Biber, S. Kannengiesser, J. Nistler, M. Braun, S. Blaess, M. Gebhardt, D. Grodzki, D. Ritter, G. Seegerer, M. Vester, and R. Schneider, “Design and operation of a whole-body MRI scanner without RF shielding,”Magnetic Resonance in Medicine, vol. 93, no. 4, pp. 1842–1855, 2025. [Online]. Available: https://onlinelibrary.wiley.com/doi/abs/10.1002/mrm.30374

work page doi:10.1002/mrm.30374 2025
[4]

MR artifacts: a review,

E. Bellon, E. Haacke, P. Coleman, D. Sacco, D. Steiger, and R. Gangarosa, “MR artifacts: a review,”American Journal of Roentgenology, vol. 147, no. 6, pp. 1271–1281, 1986, pMID: 3490763. [Online]. Available: https://doi.org/10.2214/ajr.147.6.1271

work page doi:10.2214/ajr.147.6.1271 1986
[5]

Electromagnetic Noise Characterization and Suppression in Low-Field MRI Systems,

T. Guallart-Naval, J. M. Algar ´ın, and J. Alonso, “Electromagnetic Noise Characterization and Suppression in Low-Field MRI Systems,”Magnetic Resonance in Medicine, vol. 95, no. 5, pp. 3000–3007, 2026. [Online]. Available: https://onlinelibrary.wiley.com/doi/abs/10.1002/mrm.70235

work page doi:10.1002/mrm.70235 2026
[6]

E. M. Haacke, R. W. Brown, M. R. Thompson, R. Venkatesanet al., Magnetic resonance imaging: physical principles and sequence design. Wiley-liss New York:, 1999, vol. 82

work page 1999
[7]

New methods for MRI denoising based on sparseness and self-similarity,

J. V. Manj ´on, P. Coup ´e, A. Buades, D. Louis Collins, and M. Robles, “New methods for MRI denoising based on sparseness and self-similarity,”Medical Image Analysis, vol. 16, no. 1, pp. 18–27,

work page
[8]

Available: https://www.sciencedirect.com/science/article/ pii/S1361841511000491

[Online]. Available: https://www.sciencedirect.com/science/article/ pii/S1361841511000491

work page
[9]

A survey on the magnetic resonance image denoising methods,

J. Mohan, V. Krishnaveni, and Y. Guo, “A survey on the magnetic resonance image denoising methods,”Biomedical Signal Processing and Control, vol. 9, pp. 56–69, 2014. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S1746809413001407

work page 2014
[10]

Effects of k-space spatial low-pass filtering on bio-imaging analysis,

A. Kyung, Y. J. Kwon, and L. Chung, “Effects of k-space spatial low-pass filtering on bio-imaging analysis,” in2018 9th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEM- CON), 2018, pp. 742–745

work page 2018
[11]

Well-Designed k-Space Coverage is Important for Good MRI Denoising,

J. Wang and J. P. Haldar, “Well-Designed k-Space Coverage is Important for Good MRI Denoising,”arXiv preprint arXiv:2511.05735, 2025. [Online]. Available: https://arxiv.org/abs/2511.05735

work page arXiv 2025
[12]

Two-dimensional image noise removal and reconstruction using discrete Fourier transform, k-space filtering and Z-space filtering,

C. Ciulla, “Two-dimensional image noise removal and reconstruction using discrete Fourier transform, k-space filtering and Z-space filtering,” Progress in Engineering Science, vol. 2, no. 1, p. 100056, 2025. [Online]. Available: https://www.sciencedirect.com/science/article/pii/ S2950425225000088

work page 2025
[13]

MRI denoising using deep learning,

J. V. Manj ´on and P. Coupe, “MRI denoising using deep learning,” in International workshop on patch-based techniques in medical imaging. Springer, 2018, pp. 12–19

work page 2018
[14]

Deep learning techniques for inverse problems in imaging,

G. Ongie, A. Jalal, C. A. Metzler, R. G. Baraniuk, A. G. Dimakis, and R. Willett, “Deep learning techniques for inverse problems in imaging,” IEEE Journal on Selected Areas in Information Theory, vol. 1, no. 1, pp. 39–56, 2020

work page 2020
[15]

SNRAware: Improved Deep Learning MRI Denoising with Signal-to-Noise Ratio Unit Training and G-Factor Map Augmentation,

H. Xue, S. M. Hooper, I. Pierce, R. H. Davies, J. Stairs, J. Naegele, A. E. Campbell-Washburn, C. Manisty, J. C. Moon, T. A. Treibel, M. S. Hansen, and P. Kellman, “SNRAware: Improved Deep Learning MRI Denoising with Signal-to-Noise Ratio Unit Training and G-Factor Map Augmentation,”Radiology: Artificial Intelligence, vol. 7, no. 6, p. e250227, 2025, pMID...

work page doi:10.1148/ryai.250227 2025
[16]

Imaging Transformer for MRI Denoising: a Scalable Model Architecture that enables SNR≪1 Imaging,

H. Xue, S. M. Hooper, R. H. Davies, T. A. Treibel, I. Pierce, J. Stairs, J. Naegele, C. Manisty, J. C. Moon, A. E. Campbell-Washburn, P. Kellman, and M. S. Hansen, “Imaging Transformer for MRI Denoising: a Scalable Model Architecture that enables SNR≪1 Imaging,”arXiv preprint arXiv: 2504.10534, 2025. [Online]. Available: https://arxiv.org/abs/2504.10534

work page arXiv 2025
[17]

Five steps to make MRI scanners more affordable to the world,

A. Webb and J. Obungoloch, “Five steps to make MRI scanners more affordable to the world,”Nature, vol. 615, pp. 391–393, 3 2023. [Online]. Available: https://www.nature.com/articles/d41586-023-00759-x

work page 2023
[18]

Portable magnetic resonance imaging of patients indoors, outdoors and at home,

T. Guallart-Naval, J. M. Algar ´ın, R. Pellicer-Guridi, F. Galve, Y. Vives-Gilabert, R. Bosch, E. Pall ´as, J. M. Gonz ´alez, J. P. Rigla, P. Mart´ınez, F. Lloris, J. Borreguero,´A. Marcos-Perucho, V. Negnevitsky, L. Mart ´ı-Bonmat´ı, A. R ´ıos, J. M. Benlloch, and J. Alonso, “Portable magnetic resonance imaging of patients indoors, outdoors and at home,”...

work page 2022
[19]

Whole-body magnetic resonance imaging at 0.05 tesla,

Y. Zhao, Y. Ding, V. Lau, C. Man, S. Su, L. Xiao, A. T. L. Leong, and E. X. Wu, “Whole-body magnetic resonance imaging at 0.05 tesla,”Science, vol. 384, 5 2024. [Online]. Available: https://www.science.org/doi/10.1126/science.adm7168

work page doi:10.1126/science.adm7168 2024
[20]

Low-Field MRI: How Low Can We Go? A Fresh View on an Old Debate,

M. Sarracanie and N. Salameh, “Low-Field MRI: How Low Can We Go? A Fresh View on an Old Debate,”Frontiers in Physics, vol. 8, p. 172, jun 2020. [Online]. Available: https://www.frontiersin.org/article/10.3389/fphy.2020.00172/full

work page doi:10.3389/fphy.2020.00172/full 2020
[21]

MRI at low field: A review of software solutions for improving SNR,

R. Ayde, M. Vornehm, Y. Zhao, F. Knoll, E. X. Wu, and M. Sarracanie, “MRI at low field: A review of software solutions for improving SNR,” NMR in Biomedicine, vol. 38, no. 1, p. e5268, 2025

work page 2025
[22]

Nonlocal transform-domain filter for volumetric data denoising and reconstruction,

M. Maggioni, V. Katkovnik, K. Egiazarian, and A. Foi, “Nonlocal transform-domain filter for volumetric data denoising and reconstruction,” IEEE Transactions on Image Processing, vol. 22, no. 1, pp. 119–133, 2013

work page 2013
[23]

A low-cost and shielding-free ultra-low-field brain MRI scanner,

Y. Liu, A. T. L. Leong, Y. Zhao, L. Xiao, H. K. F. Mak, A. Chun, O. Tsang, G. K. K. Lau, G. K. K. Leung, E. X. Wu, and X. Linfang, “A low-cost and shielding-free ultra-low-field brain MRI scanner,”Nature Communications 2021 12:1, vol. 12, no. 1, pp. 1–14, dec 2021. [Online]. Available: https://www.nature.com/articles/s41467-021-27317-1

work page 2021
[24]

Robust emi elimination for rf shielding-free mri through deep learning direct mr signal prediction,

Y. Zhao, L. Xiao, J. Hu, and E. X. Wu, “Robust emi elimination for rf shielding-free mri through deep learning direct mr signal prediction,”Magnetic Resonance in Medicine, vol. 92, no. 1, pp. 112–127, 2024. [Online]. Available: https://onlinelibrary.wiley.com/doi/ abs/10.1002/mrm.30046

work page doi:10.1002/mrm.30046 2024
[25]

Denoising low-field MR images with a deep learning algorithm based on simulated data from easily accessible open-source software,

A. Salehi, M. Mach, C. Najac, B. Lena, T. O’Reilly, Y. Dong, P. B ¨ornert, H. Adams, T. Evans, and A. Webb, “Denoising low-field MR images with a deep learning algorithm based on simulated data from easily accessible open-source software,”Journal of Magnetic Resonance, vol. 370, p. 107812, 2025. [Online]. Available: https://www.sciencedirect.com/science/a...

work page 2025
[27]

Available: https://arxiv.org/abs/2509.11790

[Online]. Available: https://arxiv.org/abs/2509.11790

work page arXiv
[28]

Elliptical halbach magnet and gradient modules for low-field portable magnetic resonance imaging,

F. Galve, E. Pall ´as, T. Guallart-Naval, P. Garc ´ıa-Crist´obal, P. Mart´ınez, J. M. Algar ´ın, J. Borreguero, R. Bosch, F. Juan-Lloris, J. M. Benlloch, and J. Alonso, “Elliptical halbach magnet and gradient modules for low-field portable magnetic resonance imaging,”NMR in Biomedicine, vol. 37, no. 12, p. e5258, 2024. [Online]. Available: https://analyti...

work page doi:10.1002/nbm.5258 2024
[29]

MaRCoS, an open-source electronic control system for low-field MRI,

V. Negnevitsky, Y. Vives-Gilabert, J. M. Algar ´ın, L. Craven-Brightman, 13 R. Pellicer-Guridi, T. O’Reilly, J. P. Stockmann, A. Webb, J. Alonso, and B. Menk¨ uc, “MaRCoS, an open-source electronic control system for low-field MRI,”Journal of Magnetic Resonance, vol. 350, p. 107424,

work page
[30]

Available: https://doi.org/10.1016/j.jmr.2023.107424

[Online]. Available: https://doi.org/10.1016/j.jmr.2023.107424

work page doi:10.1016/j.jmr.2023.107424 2023
[31]

Benchmarking the performance of a low-cost magnetic resonance control system at multiple sites in the open MaRCoS community,

T. Guallart-Naval, T. O’Reilly, J. M. Algar´ın, R. Pellicer-Guridi, Y. Vives- Gilabert, L. Craven-Brightman, V. Negnevitsky, B. Menk¨ uc, F. Galve, J. P. Stockmann, A. Webb, and J. Alonso, “Benchmarking the performance of a low-cost magnetic resonance control system at multiple sites in the open MaRCoS community,”NMR in Biomedicine, vol. 36, pp. 1–13, 2022

work page 2022
[32]

MaRGE: A graphical environment for MaRCoS,

J. M. Algar ´ın, T. Guallart-Naval, J. Borreguero, F. Galve, and J. Alonso, “MaRGE: A graphical environment for MaRCoS,”Journal of Magnetic Resonance, vol. 361, p. 107662, 2024

work page 2024
[33]

Github - microsoft/tyger: Remote signal processing

“Github - microsoft/tyger: Remote signal processing.” [Online]. Available: https://github.com/microsoft/tyger

work page
[34]

A fully open-source framework for streaming and cloud-processing of low-field MRI data,

T. Guallart-Naval, J. Stairs, J. M. Algar´ın, H. Xue, J. Benlloch, P. Benlloch, J. Borreguero, J. Conejero, F. Galve, P. Garc ´ıa-Crist´obal, M. Lacalle, B. Lena, L. Porcar, S. J. Schiff, A. Webb, M. Hansen, and J. Alonso, “A fully open-source framework for streaming and cloud-processing of low-field MRI data,”arXiv preprint arXiv: 2603.19287, 2026. [Onli...

work page arXiv 2026
[35]

Image reconstruction in SNR units: A general method for SNR measurement,

P. Kellman and E. R. McVeigh, “Image reconstruction in SNR units: A general method for SNR measurement,”Magnetic Resonance in Medicine, vol. 54, no. 6, pp. 1439–1447, 2005

work page 2005
[36]

Low-field musculoskeletal MRI,

S. Ghazinoor, J. V. Crues, and C. Crowley, “Low-field musculoskeletal MRI,”Journal of Magnetic Resonance Imaging, vol. 25, pp. 234–244, 2

work page
[37]

Available: http://doi.wiley.com/10.1002/jmri.20854

[Online]. Available: http://doi.wiley.com/10.1002/jmri.20854

work page doi:10.1002/jmri.20854
[38]

The Rician distribution of noisy MRI data,

H. Gudbjartsson and S. Patz, “The Rician distribution of noisy MRI data,” Magnetic resonance in medicine, vol. 34, no. 6, pp. 910–914, 1995

work page 1995
[39]

Hallucination Index: An Image Quality Metric for Generative Reconstruction Models,

M. Tivnan, S. Yoon, Z. Chen, X. Li, D. Wu, and Q. Li, “Hallucination Index: An Image Quality Metric for Generative Reconstruction Models,” inMedical Image Computing and Computer Assisted Intervention – MICCAI 2024, M. G. Linguraru, Q. Dou, A. Feragen, S. Giannarou, B. Glocker, K. Lekadir, and J. A. Schnabel, Eds. Springer Nature Switzerland, 2024, pp. 449–458

work page 2024
[40]

Zero-echo-time sequences in highly inhomogeneous fields,

J. Borreguero, F. Galve, J. M. Algar ´ın, and J. Alonso, “Zero-echo-time sequences in highly inhomogeneous fields,”Magnetic Resonance in Medicine, vol. 93, no. 3, pp. 1190–1204, 2025. [Online]. Available: https://onlinelibrary.wiley.com/doi/abs/10.1002/mrm.30352

work page doi:10.1002/mrm.30352 2025