Academic Journal

Fast zero-shot deep learning-based denoising method for low-field MR images.

Λεπτομέρειες βιβλιογραφικής εγγραφής
Τίτλος: Fast zero-shot deep learning-based denoising method for low-field MR images.
Συγγραφείς: Ayde R; Center for Adaptable MRI Technology, School of Medicine, Medical Sciences and Nutrition, University of Aberdeen, Aberdeen, UK. reina.ayde@gmail.com.; Department of Radiology, University of Michigan, Ann Arbor, USA. reina.ayde@gmail.com., Zihlmann G; Center for Adaptable MRI Technology, School of Medicine, Medical Sciences and Nutrition, University of Aberdeen, Aberdeen, UK., Salameh N; Center for Adaptable MRI Technology, School of Medicine, Medical Sciences and Nutrition, University of Aberdeen, Aberdeen, UK., Sarracanie M; Center for Adaptable MRI Technology, School of Medicine, Medical Sciences and Nutrition, University of Aberdeen, Aberdeen, UK.
Πηγή: Magma (New York, N.Y.) [MAGMA] 2026 Aug; Vol. 39 (4), pp. 531-544. Date of Electronic Publication: 2025 Dec 22.
Τύπος έκδοσης: Journal Article
Γλώσσα: English
Στοιχεία περιοδικού: Publisher: Springer Country of Publication: Germany NLM ID: 9310752 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1352-8661 (Electronic) Linking ISSN: 09685243 NLM ISO Abbreviation: MAGMA Subsets: MEDLINE
Imprint Name(s): Publication: 2003- : Heidelberg : Springer
Original Publication: New York, NY : Chapman & Hall, c1993-
Ιατρικοί όροι (MeSH): Magnetic Resonance Imaging*/methods , Image Processing, Computer-Assisted*/methods , Deep Learning*, Brain/diagnostic imaging ; Image Enhancement/methods ; Image Interpretation, Computer-Assisted/methods ; Humans ; Signal-To-Noise Ratio ; Algorithms
Περίληψη: Objective: Denoising low-field MR images is often essential to obtain image quality that is adequate for clinical diagnosis while keeping scan time patient-friendly. The recently introduced zero-shot self-supervised approach shows great promise, requiring no prior data collection for training, which is particularly challenging at low-field. Here, this scan-specific denoising approach is adapted to low-field MR data and optimized to accelerate the training process.
Material and Method: We extended the zero-shot noise-as-clean method by modifying the training process to achieve faster training times. The proposed method was compared to BM4D and the recent zero-shot noise2noise methods. Denoising performance was first evaluated quantitatively on high-field data where high SNR images are available, then assessed qualitatively on prospective low-field data (0.1 T). Ultimately, we studied the denoising performance with respect to training on portions of the original data matrix as a potential strategy for further training acceleration.
Results: The proposed method achieved high denoising performance across different SNR levels within a few seconds on a GPU for typical low-field data dimensions. Additionally, training on portion of the data showed potential for further training acceleration.
Discussion: In the context of low-field MRI, this denoising method shows great potential, as it could be integrated into acquisition workflows relatively seamlessly to improve image quality. Code: https://github.com/reinaayde7/zs-nac.git.
(© 2025. The Author(s).)
Competing Interests: Declarations. Competing interest: The authors have no competing interests to declare that are relevant to the content of this article. Ethical approval: All MRI experiments were conducted following the local ethics regulations and informed consent was obtained from all subjects. The study was approved by the Ethikkommission Nordwest- und Zentralschweiz (EKNZ) (project-ID 2022-00348) and the University of Aberdeen School Ethics Review Board (School of Medicine, Medical Sciences and Nutrition, REC reference number 2144417).
References: Nat Mach Intell. 2022;4(11):953-963. (PMID: 36415333)
IEEE Trans Med Imaging. 2020 Apr;39(4):1064-1072. (PMID: 31535985)
Invest Radiol. 2022 Apr 1;57(4):263-271. (PMID: 34669651)
IEEE Trans Image Process. 2020 Sep 30;PP:. (PMID: 32997627)
Sci Rep. 2023 Dec 1;13(1):21183. (PMID: 38040835)
Front Artif Intell. 2021 Jun 04;4:642731. (PMID: 34151253)
MAGMA. 2024 Jul;37(3):335-368. (PMID: 39042206)
Magn Reson Med. 1990 Nov;16(2):192-225. (PMID: 2266841)
IEEE Trans Image Process. 2013 Jan;22(1):119-33. (PMID: 22868570)
J Magn Reson. 2025 Jan;370:107812. (PMID: 39647413)
Sci Rep. 2021 Apr 15;11(1):8248. (PMID: 33859218)
Front Neuroimaging. 2025 May 06;4:1501801. (PMID: 40395481)
J Magn Reson. 2022 Nov;344:107319. (PMID: 36332511)
Radiol Artif Intell. 2020 Jan 29;2(1):e190007. (PMID: 32076662)
NMR Biomed. 2025 Jan;38(1):e5268. (PMID: 39375036)
Sci Rep. 2022 Jul 30;12(1):13147. (PMID: 35907975)
IEEE Trans Image Process. 2007 Aug;16(8):2080-95. (PMID: 17688213)
Magn Reson Med. 2014 Mar;71(3):990-1001. (PMID: 23649942)
Magn Reson Med. 1995 Dec;34(6):910-4. (PMID: 8598820)
Contributed Indexing: Keywords: Deep learning; Denoising; Low-field; MRI; Self-supervised; Zero-shot
Entry Date(s): Date Created: 20251222 Date Completed: 20260807 Latest Revision: 20260813
Update Code: 20260813
PubMed Central ID: PMC13447554
DOI: 10.1007/s10334-025-01311-w
PMID: 41428276
Βάση Δεδομένων: MEDLINE
Περιγραφή
ISSN:1352-8661
DOI:10.1007/s10334-025-01311-w