NAF-GAN: Anatomically Constrained GAN for Ultra-Low-Field MRI Enhancement

Zhenyu Xiang; Yicheng Wu; Ziyang Chen; Zaiyuan Liu; Yongsheng Pan; Yong Xia

doi:10.1007/978-3-032-23344-8_11

NAF-GAN: Anatomically Constrained GAN for Ultra-Low-Field MRI Enhancement

Zhenyu Xiang
, Yicheng Wu
, Ziyang Chen
, Zaiyuan Liu
, Yongsheng Pan
, Yong Xia

School of Computer Science

Northwestern Polytechnical University Xian
Monash University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

High-quality diagnostic imaging in resource-limited settings remains a significant challenge due to the limited accessibility of high-field Magnetic Resonance Imaging (MRI). The Ultra-Low-Field MRI Image Enhancement Challenge (ULF-EnC) addresses this issue by leveraging paired 64mT and 3T MRI scans to develop algorithms that enhance low-field image quality to high-field standards. In this paper, we propose an end-to-end NAF-GAN framework, a generative adversarial network based on the Nonlinear Activation Free Network (NAFNet) architecture, specifically designed for this task. Our model learns the complex translation from 64mT to 3T MRI sequences, improving image clarity and contrast while preserving the anatomical fidelity essential for clinical diagnosis. Both the patch-level adversarial loss and the segmentation-based Dice loss are designed to capture high-frequency details and preserve anatomical structures. The proposed approach achieves strong performance across T1, T2, and FLAIR contrasts, underscoring its potential to enable cost-effective and accessible MRI solutions for mobile and global healthcare applications.

Original language	English
Title of host publication	Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging - 1st International Challenge, ULF-EnC 2025, Held in Conjunction with MICCAI 2025, Proceedings
Editors	Zhaolin Chen, Sanuwani Dayarathna, Kh Tohidul Islam, Himashi Peiris, Parisa Zakavi, Shenjun Zhong
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	94-103
Number of pages	10
ISBN (Print)	9783032233431
DOIs	https://doi.org/10.1007/978-3-032-23344-8_11
State	Published - 2026
Event	1st ULF-EnC 2025 Challenge on Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging, held in conjunction with MICCAI 2025 - Daejeon, Korea, Republic of Duration: 23 Sep 2025 → 23 Sep 2025

Publication series

Name	Lecture Notes in Computer Science
Volume	16293 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	1st ULF-EnC 2025 Challenge on Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging, held in conjunction with MICCAI 2025
Country/Territory	Korea, Republic of
City	Daejeon
Period	23/09/25 → 23/09/25

Keywords

Anatomical Structures
NAFNet
Ultra-Low-Field MRI

Access to Document

10.1007/978-3-032-23344-8_11

Cite this

Xiang, Z., Wu, Y., Chen, Z., Liu, Z., Pan, Y., & Xia, Y. (2026). NAF-GAN: Anatomically Constrained GAN for Ultra-Low-Field MRI Enhancement. In Z. Chen, S. Dayarathna, K. T. Islam, H. Peiris, P. Zakavi, & S. Zhong (Eds.), Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging - 1st International Challenge, ULF-EnC 2025, Held in Conjunction with MICCAI 2025, Proceedings (pp. 94-103). (Lecture Notes in Computer Science; Vol. 16293 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-032-23344-8_11

Xiang, Zhenyu ; Wu, Yicheng ; Chen, Ziyang et al. / NAF-GAN : Anatomically Constrained GAN for Ultra-Low-Field MRI Enhancement. Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging - 1st International Challenge, ULF-EnC 2025, Held in Conjunction with MICCAI 2025, Proceedings. editor / Zhaolin Chen ; Sanuwani Dayarathna ; Kh Tohidul Islam ; Himashi Peiris ; Parisa Zakavi ; Shenjun Zhong. Springer Science and Business Media Deutschland GmbH, 2026. pp. 94-103 (Lecture Notes in Computer Science).

@inproceedings{028371042a57403ba0f8afe0b71076c3,

title = "NAF-GAN: Anatomically Constrained GAN for Ultra-Low-Field MRI Enhancement",

abstract = "High-quality diagnostic imaging in resource-limited settings remains a significant challenge due to the limited accessibility of high-field Magnetic Resonance Imaging (MRI). The Ultra-Low-Field MRI Image Enhancement Challenge (ULF-EnC) addresses this issue by leveraging paired 64mT and 3T MRI scans to develop algorithms that enhance low-field image quality to high-field standards. In this paper, we propose an end-to-end NAF-GAN framework, a generative adversarial network based on the Nonlinear Activation Free Network (NAFNet) architecture, specifically designed for this task. Our model learns the complex translation from 64mT to 3T MRI sequences, improving image clarity and contrast while preserving the anatomical fidelity essential for clinical diagnosis. Both the patch-level adversarial loss and the segmentation-based Dice loss are designed to capture high-frequency details and preserve anatomical structures. The proposed approach achieves strong performance across T1, T2, and FLAIR contrasts, underscoring its potential to enable cost-effective and accessible MRI solutions for mobile and global healthcare applications.",

keywords = "Anatomical Structures, NAFNet, Ultra-Low-Field MRI",

author = "Zhenyu Xiang and Yicheng Wu and Ziyang Chen and Zaiyuan Liu and Yongsheng Pan and Yong Xia",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.; 1st ULF-EnC 2025 Challenge on Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging, held in conjunction with MICCAI 2025 ; Conference date: 23-09-2025 Through 23-09-2025",

year = "2026",

doi = "10.1007/978-3-032-23344-8\_11",

language = "英语",

isbn = "9783032233431",

series = "Lecture Notes in Computer Science",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "94--103",

editor = "Zhaolin Chen and Sanuwani Dayarathna and Islam, \{Kh Tohidul\} and Himashi Peiris and Parisa Zakavi and Shenjun Zhong",

booktitle = "Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging - 1st International Challenge, ULF-EnC 2025, Held in Conjunction with MICCAI 2025, Proceedings",

}

Xiang, Z, Wu, Y, Chen, Z, Liu, Z, Pan, Y & Xia, Y 2026, NAF-GAN: Anatomically Constrained GAN for Ultra-Low-Field MRI Enhancement. in Z Chen, S Dayarathna, KT Islam, H Peiris, P Zakavi & S Zhong (eds), Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging - 1st International Challenge, ULF-EnC 2025, Held in Conjunction with MICCAI 2025, Proceedings. Lecture Notes in Computer Science, vol. 16293 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 94-103, 1st ULF-EnC 2025 Challenge on Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging, held in conjunction with MICCAI 2025, Daejeon, Korea, Republic of, 23/09/25. https://doi.org/10.1007/978-3-032-23344-8_11

NAF-GAN: Anatomically Constrained GAN for Ultra-Low-Field MRI Enhancement. / Xiang, Zhenyu; Wu, Yicheng; Chen, Ziyang et al.
Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging - 1st International Challenge, ULF-EnC 2025, Held in Conjunction with MICCAI 2025, Proceedings. ed. / Zhaolin Chen; Sanuwani Dayarathna; Kh Tohidul Islam; Himashi Peiris; Parisa Zakavi; Shenjun Zhong. Springer Science and Business Media Deutschland GmbH, 2026. p. 94-103 (Lecture Notes in Computer Science; Vol. 16293 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - NAF-GAN

T2 - 1st ULF-EnC 2025 Challenge on Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging, held in conjunction with MICCAI 2025

AU - Xiang, Zhenyu

AU - Wu, Yicheng

AU - Chen, Ziyang

AU - Liu, Zaiyuan

AU - Pan, Yongsheng

AU - Xia, Yong

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.

PY - 2026

Y1 - 2026

N2 - High-quality diagnostic imaging in resource-limited settings remains a significant challenge due to the limited accessibility of high-field Magnetic Resonance Imaging (MRI). The Ultra-Low-Field MRI Image Enhancement Challenge (ULF-EnC) addresses this issue by leveraging paired 64mT and 3T MRI scans to develop algorithms that enhance low-field image quality to high-field standards. In this paper, we propose an end-to-end NAF-GAN framework, a generative adversarial network based on the Nonlinear Activation Free Network (NAFNet) architecture, specifically designed for this task. Our model learns the complex translation from 64mT to 3T MRI sequences, improving image clarity and contrast while preserving the anatomical fidelity essential for clinical diagnosis. Both the patch-level adversarial loss and the segmentation-based Dice loss are designed to capture high-frequency details and preserve anatomical structures. The proposed approach achieves strong performance across T1, T2, and FLAIR contrasts, underscoring its potential to enable cost-effective and accessible MRI solutions for mobile and global healthcare applications.

AB - High-quality diagnostic imaging in resource-limited settings remains a significant challenge due to the limited accessibility of high-field Magnetic Resonance Imaging (MRI). The Ultra-Low-Field MRI Image Enhancement Challenge (ULF-EnC) addresses this issue by leveraging paired 64mT and 3T MRI scans to develop algorithms that enhance low-field image quality to high-field standards. In this paper, we propose an end-to-end NAF-GAN framework, a generative adversarial network based on the Nonlinear Activation Free Network (NAFNet) architecture, specifically designed for this task. Our model learns the complex translation from 64mT to 3T MRI sequences, improving image clarity and contrast while preserving the anatomical fidelity essential for clinical diagnosis. Both the patch-level adversarial loss and the segmentation-based Dice loss are designed to capture high-frequency details and preserve anatomical structures. The proposed approach achieves strong performance across T1, T2, and FLAIR contrasts, underscoring its potential to enable cost-effective and accessible MRI solutions for mobile and global healthcare applications.

KW - Anatomical Structures

KW - NAFNet

KW - Ultra-Low-Field MRI

UR - https://www.scopus.com/pages/publications/105043998854

U2 - 10.1007/978-3-032-23344-8_11

DO - 10.1007/978-3-032-23344-8_11

M3 - 会议稿件

AN - SCOPUS:105043998854

SN - 9783032233431

T3 - Lecture Notes in Computer Science

SP - 94

EP - 103

BT - Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging - 1st International Challenge, ULF-EnC 2025, Held in Conjunction with MICCAI 2025, Proceedings

A2 - Chen, Zhaolin

A2 - Dayarathna, Sanuwani

A2 - Islam, Kh Tohidul

A2 - Peiris, Himashi

A2 - Zakavi, Parisa

A2 - Zhong, Shenjun

PB - Springer Science and Business Media Deutschland GmbH

Y2 - 23 September 2025 through 23 September 2025

ER -

Xiang Z, Wu Y, Chen Z, Liu Z, Pan Y, Xia Y. NAF-GAN: Anatomically Constrained GAN for Ultra-Low-Field MRI Enhancement. In Chen Z, Dayarathna S, Islam KT, Peiris H, Zakavi P, Zhong S, editors, Enhancing Ultra-Low-Field MRI with Paired High-Field MRI Comparisons for Brain Imaging - 1st International Challenge, ULF-EnC 2025, Held in Conjunction with MICCAI 2025, Proceedings. Springer Science and Business Media Deutschland GmbH. 2026. p. 94-103. (Lecture Notes in Computer Science). doi: 10.1007/978-3-032-23344-8_11