OSCR

Time-Conditioned Zero-Shot Self-Supervised Reconstruction for Accelerated 3D Ultra-Low-Field MRI.

Code ↔ Paper

8 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 8 matches · 2 of them tie a paragraph to a whole file, not to given lines: weak matches, whose lines are not tinted
  1. [1] § Methods › Retrospective Undersampling Experiments ↔ codes/mask_generator.py, lines 93–124 · score 0.77 · Poisson disc, calibration region, phase encoding, masks, acceleration, undersampling
  2. [2] § Theory › Zero‐Shot Self‐Supervised Learning (ZS‐SSL) ↔ codes/parser_ops.py, the whole file · a weak match · score 0.70 · conjugate gradient, zero shot, residual blocks, supervised learning, validation mask, ResNet
  3. [3] § Methods › Model Architecture, Hyperparameter Tuning, and Pretraining ↔ codes/modules.py, lines 43–126 · score 0.64 · random affine transformations, random flips
  4. [4] § Methods › Model Architecture, Hyperparameter Tuning, and Pretraining ↔ codes/parser_ops.py, the whole file · a weak match · score 0.62 · conjugate gradient iterations, residual blocks, ResNet, ZS SSL, Hyperparameter, slice
  5. [5] § Methods › Model Architecture, Hyperparameter Tuning, and Pretraining ↔ ulf_3d_zs_ssl_recon.ipynb, lines 124–143 · score 0.61 · L1 L2 loss, Adam, Mixed, blocks, optimizer, unrolled
  6. [6] § Methods › Model Architecture, Hyperparameter Tuning, and Pretraining ↔ codes/model_3d.py, lines 31–55 · score 0.57 · residual scaling, ResNet, Sinusoidal, PyTorch, embeddings, blocks
  7. [7] § Theory › Zero‐Shot Self‐Supervised Learning (ZS‐SSL) ↔ codes/model_3d.py, lines 183–213 · score 0.55 · unrolled optimization, ResNet, loss mask, module, DC, blocks
  8. [8] § Results ↔ ulf_3d_zs_ssl_recon.ipynb, lines 124–143 · score 0.55 · pretrained weights, transfer learning, TL model, ZS SSL, loss, ULF

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

Python · 47 lines · 2.8 KB · no license · 2 matches

  1. import argparse
  3. def get_parser():
  4. parser = argparse.ArgumentParser(description='ZS-SSL: Zero-Shot Self-Supervised Learning')
  6. # Hyperparameters for the network
  7. parser.add_argument('--data_dir', type=str, default='data/brain_T1w.npy',
  8. help='data directory')
  9. parser.add_argument('--ncontrast', type=int, default=1,
  10. help='number of contrasts of the slices in the dataset')
  11. parser.add_argument('--ndepth', type=int, default=38,
  12. help='number of slices in the dataset')
  13. parser.add_argument('--nrow', type=int, default=150,
  14. help='number of rows of the slices in the dataset')
  15. parser.add_argument('--ncol', type=int, default=136,
  16. help='number of columns of the slices in the dataset')
  17. parser.add_argument('--ncoil', type=int, default=1,
  18. help='number of coils of the slices in the dataset')
  19. parser.add_argument('--acc_rate', type=int, default=2,
  20. help='acceleration rate')
  21. parser.add_argument('--epochs', type=int, default=100,
  22. help='number of epochs to train')
  23. parser.add_argument('--learning_rate', type=float, default=5e-4,
  24. help='learning rate')
  25. parser.add_argument('--batchSize', type=int, default=1,
  26. help='batch size')
  27. parser.add_argument('--nb_unroll_blocks', type=int, default=5,
  28. help='number of unrolled blocks')
  29. parser.add_argument('--nb_res_blocks', type=int, default=5,
  30. help="number of residual blocks in ResNet")
  31. parser.add_argument('--CG_Iter', type=int, default=10,
  32. help='number of Conjugate Gradient iterations for DC')
  34. # Hyperparameters for the 3D-ZS-SSL
  35. parser.add_argument('--rho_val', type=float, default=0.2,
  36. help='cardinality of the validation mask')
  37. parser.add_argument('--rho_train', type=float, default=0.4,
  38. help='cardinality of the loss mask, \ rho = |\ Lambda| / |\ Omega|')
  39. parser.add_argument('--num_reps', type=int, default=25,
  40. help='number of repetions for the remainder mask')
  41. parser.add_argument('--transfer_learning', type=bool, default=False,
  42. help='transfer learning from pretrained model')
  43. parser.add_argument('--TL_path', type=str, default="pretrained_weights/Pretraining_R2_5Unrolls_5ResNet/best.pth",
  44. help='path to pretrained model')
  45. parser.add_argument('--stop_training', type=int, default=5,
  46. help='stop training if a new lowest validation loss hasnt been achieved in xx epochs')
  47. return parser

parser_ops.py at commit 7009dcf, no license · at the source

Overview

Authors: Mart W J van Straten1,2, Beatrice Lena1, Chloé Najac1, Ruben van den Broek1, Peter Börnert1,3, Andrew Webb1, Yiming Dong1
ORCID iDs: Mart W J van Straten, Beatrice Lena, Chloé Najac, Yiming Dong
  1. C.J. Gorter MRI Center, Department of Radiology, LUMC, Leiden, the Netherlands
  2. Department of Biomedical Engineering, Eindhoven University of Technology, Eindhoven, the Netherlands
  3. Philips Innovative Technologies Hamburg, Hamburg, Germany
Institutions: Leiden University Medical Center (Netherlands); Eindhoven University of Technology (Netherlands); Philips (Germany) (Germany)
Journal: Magnetic resonance in medicine, volume 96, issue 3, pages 1303-1312
Dates: received 19 December 2025; accepted 16 April 2026; published online 28 April 2026; in print September 2026
Type: Other · Language: English
License: CC BY
Identifiers: DOI 10.1002/mrm.70407 · PMID 42050898 · PMCID PMC13327438 · OpenAlex W7158035497
Open access: hybrid, a free copy (OpenAlex)
Status: code verified
Categories: structural MRI / diffusion (modality), human (organism)
Methods: Machine learning
Keywords: 3D image reconstruction, physics‐guided deep learning, ultra‐low‐field MRI, undersampled MRI, zero‐shot self‐supervised learning
MeSH: Brain*, Image Processing, Computer-Assisted*, Imaging, Three-Dimensional*, Magnetic Resonance Imaging*, Supervised Machine Learning*, Algorithms, Humans, Signal-To-Noise Ratio, Transfer Machine Learning (* major topic)
Topic: Advanced MRI Techniques and Applications (Radiology, Nuclear Medicine and Imaging, Medicine), according to OpenAlex
Funding: Dutch Research Council (NWO) (18981); Dutch Science Foundation (NWO) Open Technology Grant (18981); European Research Council (101021218)
Citations: not cited yet (Europe PMC); 26 references in the paper

Abstract

Purpose: Ultra‐low‐field (ULF) MRI provides a cost‐effective, portable imaging option but has relatively low SNR and long acquisition times compared to standard clinical scans. This study presents a time‐conditioned zero‐shot self‐supervised learning image reconstruction framework (ULF‐ZS‐SSL) to accelerate 3D‐acquired single‐coil ULF MRI without relying on external training data. In addition, for faster computation, a transfer‐learning (TL) variant (ULF‐ZS‐SSL‐TL) was implemented by pretraining on a small fully‐sampled ULF brain dataset and fine‐tuning on the target subject in a zero‐shot manner.

Methods: This image reconstruction method combines a physics‐based data‐consistency step with a 3D residual network prior and sinusoidal time‐step embeddings to improve convergence speed. Data were acquired on a 47 mT Halbach‐based scanner using 3D turbo spin‐echo sequences with T1‐, T2‐, and inversion‐recovery–T1‐weighted contrasts. Additional T1‐weighted wrist scans were acquired to evaluate cross‐anatomy generalization. Both true and retrospectively undersampled data were compared with total variation (TV) and model‐based deep learning (MoDL).

Results: The ULF‐ZS‐SSL method produced high‐quality reconstructions across all tested contrasts, outperforming zero‐filled and TV reconstructions, particularly at higher acceleration factors. Time‐step conditioning improved convergence speed, while ULF‐ZS‐SSL‐TL further accelerated the image reconstruction three‐fold, enabling full 3D reconstructions in about 3 min. Pretraining on brain data also worked well for wrist reconstructions, indicating cross‐anatomy generalization.

Conclusion: The ULF‐ZS‐SSL framework enables accurate, training‐free reconstruction of undersampled single‐coil ULF MRI data, as does the ULF‐ZS‐SSL‐TL approach using minimal training data. The combination of physics‐based unrolling, time‐step conditioning, and transfer‐learning supports rapid and robust application in portable or resource‐limited ULF MRI systems.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 8 matches between paragraphs and lines of code.

MartvStraten/ULF_3D_ZS_SSL

License: none: the authors keep all their rights
State: the link answers, verified on 30 September 2026
Evidence: files inventoried
Commit: 7009dcf6f653c25f13d16dca01f1194be947a712, 8 July 2026
Languages: Python (5), Jupyter (2)
Size: 18 files, 7 scripts
Software Heritage: not archived
Found in: “Data Availability Statement”
Holds: README, environment (requirements.txt), 2 notebooks
Not found: license file, CITATION.cff, tests, continuous integration, documentation
Tools: NumPy (5 files), PyTorch (5 files), Matplotlib (2 files), SciPy (2 files), scikit-image (1 file)
Availability: 1 check, the latest on 30 September 2026: the link answers
  • 30 September 2026: the link answers
8 files
At the source: github.com/MartvStraten/ULF_3D_ZS_SSL

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

  • 1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 7 scripts, each with its path and the digest of its content;
  • 8 matches between paragraphs of the paper and lines of the code (method lexical-v1);
  • neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data Availability Statement

To support reproducible research, the source code used in this study, along with an ultra‐low‐field human brain dataset, has been made publicly available. The code and data can be found in the following repository: https://github.com/MartvStraten/ULF_3D_ZS_SSL.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 30 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 7 authors, 5 keywords, 9 MeSH terms, 3 funders, 23 references.

Cite

This paper

van Straten, M. W. J., Lena, B., Najac, C., van den Broek, R., Börnert, P., Webb, A., & Dong, Y. (2026). Time-Conditioned Zero-Shot Self-Supervised Reconstruction for Accelerated 3D Ultra-Low-Field MRI. Magnetic resonance in medicine, 96(3), 1303-1312. https://doi.org/10.1002/mrm.70407

BibTeX

@article{vanstraten2026time,
author = {van Straten, Mart W J and Lena, Beatrice and Najac, Chloé and van den Broek, Ruben and Börnert, Peter and Webb, Andrew and Dong, Yiming},
title = {{Time-Conditioned Zero-Shot Self-Supervised Reconstruction for Accelerated 3D Ultra-Low-Field MRI}},
journal = {Magnetic resonance in medicine},
year = {2026},
month = apr,
volume = {96},
number = {3},
pages = {1303--1312},
publisher = {Wiley},
issn = {0740-3194},
doi = {10.1002/mrm.70407},
url = {https://doi.org/10.1002/mrm.70407},
pmid = {42050898},
pmcid = {PMC13327438}
}

RIS

TY - JOUR
AU - van Straten, Mart W J
AU - Lena, Beatrice
AU - Najac, Chloé
AU - van den Broek, Ruben
AU - Börnert, Peter
AU - Webb, Andrew
AU - Dong, Yiming
TI - Time-Conditioned Zero-Shot Self-Supervised Reconstruction for Accelerated 3D Ultra-Low-Field MRI
T2 - Magnetic resonance in medicine
J2 - Magn Reson Med
PY - 2026
DA - 2026/04/28
VL - 96
IS - 3
SP - 1303
EP - 1312
SN - 0740-3194
PB - Wiley
DO - 10.1002/mrm.70407
UR - https://doi.org/10.1002/mrm.70407
LA - en
ER -

CSL-JSON

{
"id": "10.1002/mrm.70407",
"type": "article-journal",
"title": "Time-Conditioned Zero-Shot Self-Supervised Reconstruction for Accelerated 3D Ultra-Low-Field MRI",
"container-title": "Magnetic resonance in medicine",
"author": [
{
"family": "van Straten",
"given": "Mart W J"
},
{
"family": "Lena",
"given": "Beatrice"
},
{
"family": "Najac",
"given": "Chloé"
},
{
"family": "van den Broek",
"given": "Ruben"
},
{
"family": "Börnert",
"given": "Peter"
},
{
"family": "Webb",
"given": "Andrew"
},
{
"family": "Dong",
"given": "Yiming"
}
],
"container-title-short": "Magn Reson Med",
"volume": "96",
"issue": "3",
"page": "1303-1312",
"DOI": "10.1002/mrm.70407",
"PMID": "42050898",
"PMCID": "PMC13327438",
"ISSN": "0740-3194",
"publisher": "Wiley",
"URL": "https://doi.org/10.1002/mrm.70407",
"language": "en",
"issued": {
"date-parts": [
[
2026,
4,
28
]
]
}
}

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1002/mrm.70488 [code]
Advancing the Volumetric Analysis of Ultra-Low-Field Brain MRI Using Image-to-Image Translation.
Journal: Magnetic resonance in medicine
In common: PyTorch, SciPy, Matplotlib, 1 other tool, structural MRI / diffusion, 6 references
[2] doi:10.1002/hbm.70547 [code]
miniMORPH: A Morphometry Pipeline for Low-Field MRI in Infants.
Journal: Human brain mapping
In common: Matplotlib, NumPy, structural MRI / diffusion, 3 references
[3] doi:10.1080/01652176.2026.2696026
Canine brain imaging with a new low-field portable (0.05 T) MRI scanner: a pilot <i>in vivo</i> comparison to conventional 1.5 T.
Journal: The veterinary quarterly
In common: structural MRI / diffusion, 4 references
[4] doi:10.1002/mrm.70439 [code]
Group-Patch Joint Compression: Compressing Dynamic B<sub>0</sub> and Static RF Spatial Modulations Across k-Space Subregion Groups for Highly Accelerated MRI.
Journal: Magnetic resonance in medicine
In common: structural MRI / diffusion, 4 references
[5] doi:10.1002/nbm.70368 [code]
Intra-MRI Head Motion Tracking and Correction: A Quantitative In Vivo Evaluation Framework.
Journal: NMR in biomedicine
In common: scikit-image, PyTorch, SciPy, 2 other tools, structural MRI / diffusion, 1 reference
[6] doi:10.1016/j.patter.2026.101538 [code]
A multi-modal foundation model for brain disease diagnosis and medical imaging.
Journal: Patterns (New York, N.Y.)
In common: scikit-image, PyTorch, SciPy, 2 other tools, 1 reference
[7] doi:10.1002/nbm.70305 [code]
Reducing Noise Induced by Cardiac Pulsatility in Brain Maps of R<sub>2</sub>* and Magnetic Susceptibility Using Tailored k-space Sampling.
Journal: NMR in biomedicine
In common: NumPy, structural MRI / diffusion, 3 references
[8] doi:10.1038/s41597-026-07144-z [code]
Brain MRI Dataset Featuring a Full Clinical Protocol With and Without Intentional Motion.
Journal: Scientific data
In common: scikit-image, SciPy, Matplotlib, 1 other tool, structural MRI / diffusion, 1 reference
[9] doi:10.1016/j.isci.2026.117446 [code]
Volume-inflation registration (INFREG) for morphometric analysis of human focal cortical dysplasia type II.
Journal: iScience
In common: scikit-image, PyTorch, SciPy, 2 other tools, structural MRI / diffusion
[10] doi:10.1002/hipo.70124 [code]
Association Between Anterior Hippocampal Gyrification and Episodic Memory Performance in Neurotypical Young Adults.
Journal: Hippocampus
In common: scikit-image, PyTorch, SciPy, 2 other tools, structural MRI / diffusion

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.