超极化129Xe MRI的多模态增强去噪扩散模型研究

doi:10.11938/cjmr20253153

波谱学杂志 ›› 2025, Vol. 42 ›› Issue (4): 364-377.doi: 10.11938/cjmr20253153cstr: 32225.14.cjmr20253153

超极化¹²⁹Xe MRI的多模态增强去噪扩散模型研究

张明玉¹^,³^,^#, 肖洒¹^,²^,^#, 石胜杰¹, 张学成¹, 周欣¹^,²^,³^,^*()

1.磁共振波谱与成像全国重点实验室，武汉磁共振中心，中国科学院精密测量科学与技术创新研究院，湖北武汉 430071
2.中国科学院大学，北京 100049
3.海南省生物医学工程重点实验室，海南大学生物医学工程学院，海南海口 570228

收稿日期:2025-03-26 出版日期:2025-12-05 在线发表日期:2025-06-16
通讯作者: * Tel: 027-87198631, E-mail: xinzhou@wipm.ac.cn.
作者简介:# 共同第一作者
基金资助:
国家自然科学基金(82127802);国家自然科学基金(21921004);中国科学院战略性先导研究计划(XDB0540000);中国科学院战略性先导研究计划(XDC0170000);湖北省重点技术基金(2021ACA013);湖北省尖刀攻关工程专项(2023BAA021);湖北省自然科学基金(2023AFB1061)

Research on a Multi-modal Enhanced Denoising Diffusion Model for Hyperpolarized ¹²⁹Xe MRI

ZHANG Mingyu¹^,³^,^#, XIAO Sa¹^,²^,^#, SHI Shengjie¹, ZHANG Xuecheng¹, ZHOU Xin¹^,²^,³^,^*()

1. State Key Laboratory of Magnetic Resonance Spectroscopy and Imaging, National Center for Magnetic Resonance in Wuhan, Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences, Wuhan 430071, China
2. University of Chinese Academy of Sciences, Beijing 100049, China
3. Key Laboratory of Biomedical Engineering of Hainan Province, School of Biomedical Engineering, Hainan University, Haikou 570228, China

Received:2025-03-26 Published:2025-12-05 Online:2025-06-16
Contact: * Tel: 027-87198631, E-mail: xinzhou@wipm.ac.cn.

摘要/Abstract

摘要：

超极化¹²⁹Xe MRI作为一种新兴的磁共振成像（MRI）技术，在多种肺部疾病的诊疗中发挥了重要作用．然而在采集过程中产生的噪声会对数据质量造成影响，降低了该技术在临床应用中的可靠性．为此，本文提出了一种基于深度学习的多模态特征增强条件扩散模型，以实现去除噪声，提升图像质量的目的．该模型通过输入相同屏气状态下采集的¹H MRI作为约束条件，设计了一种多模态特征增强模块，用于提高模型对多模态信息的利用和对微小局部变换的敏感性．实验结果表明，在与其他方法对比中，本文方法具有更好的去噪性能和更强的图像细节保留能力．在通气缺陷分割任务中，分割结果进一步证明了本文方法对提升¹²⁹Xe MRI临床可靠性的作用．

关键词: 扩散模型, 超极化¹²⁹Xe, 多模态, 图像去噪, 深度学习

Abstract:

Hyperpolarized ¹²⁹Xe magnetic resonance imaging (MRI) is an emerging medical imaging technique that plays an important role in the diagnosis and treatment of numerous lung diseases. However, the noise generated during the acquisition process affects the data quality and limits the reliability of the technique in clinical diagnosis and treatment. In this paper, we propose a multimodal feature-enhanced conditional diffusion model based on deep learning that aims to remove noise and improve image quality. The model inputs acquired ¹H MRI as constraints, and a multimodal feature enhancement module is specially designed, which aims to enhance the effectiveness of the model in exploiting multimodal information and the sensitivity to changes in local details of the image. The experimental results show that the method has the best denoising performance and detail preservation compared to other methods, and demonstrate in a ventilation defect segmentation task that the method can enhance the reliability of ¹²⁹Xe MRI in clinical practice.

Key words: diffusion model, hyperpolarized ¹²⁹Xe MRI, multi-modal, image denoising, deep learning

中图分类号:

O482.53

张明玉, 肖洒, 石胜杰, 张学成, 周欣. 超极化¹²⁹Xe MRI的多模态增强去噪扩散模型研究[J]. 波谱学杂志, 2025, 42(4): 364-377.

ZHANG Mingyu, XIAO Sa, SHI Shengjie, ZHANG Xuecheng, ZHOU Xin. Research on a Multi-modal Enhanced Denoising Diffusion Model for Hyperpolarized ¹²⁹Xe MRI[J]. Chinese Journal of Magnetic Resonance, 2025, 42(4): 364-377.

图/表 12

图1

图2

图3

图4

图5

图6

表1

图7

表2

图8

图9

图10

参考文献 31

[1]	KERN A L, VOGEL-CLAUSSEN J. Hyperpolarized gas MRI in pulmonology[J]. Brit J Radiol, 2018, 91(1084): 20170647.
[2]	LI H D, ZHANG Z Y, HAN Y Q, et al. Hyperpolarized gas magnetic resonance imaging of the lung[J]. Chinese J Magn Reson, 2014, 31(3): 307-320.
	李海东, 张智颖, 韩叶清, 等. 超极化气体肺部磁共振成像[J]. 波谱学杂志, 2014, 31(3): 307-320.
[3]	WANG G X, YANG H Y, LI J, et al. Overview and progress of X-nuclei magnetic resonance imaging in biomedical studies[J]. Magn Reson Lett, 2023, 3(4): 327-343.
[4]	LI H D, ZHAO X C, WANG Y J, et al. Damaged lung gas exchange function of discharged COVID-19 patients detected by hyperpolarized ¹²⁹Xe MRI[J]. Sci Adv, 2021, 7(1): eabc8180.
[5]	ZHOU Q, RAO Q, LI H D, et al. Evaluation of injuries caused by coronavirus disease 2019 using multi-nuclei magnetic resonance imaging[J]. Magn Reson Lett, 2021, 1(1): 2-10. doi: 10.1016/j.mrl.2021.100009 pmid: 35673615
[6]	MACOVSKI A. Noise in MRI[J]. Magn Reson Med, 1996, 36(3): 494-497. pmid: 8875425
[7]	MOHAN J, KRISHNAVENI V, GUO Y. A survey on the magnetic resonance image denoising methods[J]. Biomed Signal Process Control, 2014, 9(1): 56-69. doi: 10.1016/j.bspc.2013.10.007
[8]	MANJÓN J V, CARBONELL-CABALLERO J, LULL J J, et al. MRI denoising using Non-Local Means[J]. Med Image Anal, 2008, 12(4): 514-523. doi: 10.1016/j.media.2008.02.004 pmid: 18381247
[9]	MANJÓN J V, COUPE P, BUADES A, et al. New methods for MRI denoising based on sparseness and self-similarity[J]. Med Image Anal, 2012, 16(1): 18-27. doi: 10.1016/j.media.2011.04.003 pmid: 21570894
[10]	SODERLUND S A, BDAIWEI A S, PLUMMER J W, et al. Improved diffusion-weighted hyperpolarized ¹²⁹Xe lung MRI with patch-based higher-order, singular value decomposition denoising[J]. Acad Radiol, 2024, 31(12): 5289-5299. doi: 10.1016/j.acra.2024.06.029
[11]	SHI S J, WANG C, XIAO S, et al. Magnetic resonance image denoising for Rician noise using a novel hybrid transformer-CNN network (HTC-net) and self-supervised pretraining[J]. Med Phys, 2025, 52: 1643-1660. doi: 10.1002/mp.17562 pmid: 39641989
[12]	XU Y, HAN K, ZHOU Y M, et al. Deep adaptive blending network for 3D magnetic resonance image denoising[J]. IEEE J Biomed Health Inform, 2021, 25(9): 3321-3331. doi: 10.1109/JBHI.2021.3087407
[13]	DOLZ J, GOPINATH K, YUAN J, et al. HyperDense-Net: a hyper-densely connected CNN for multi-modal image segmentation[J]. IEEE Trans Med Imaging, 2018, 38(5): 1116-1126. doi: 10.1109/TMI.42
[14]	ZHANG K, ZUO W M, CHEN Y J, et al. Beyond a gaussian denoiser: residual learning of deep CNN for image denoising[J]. IEEE Trans Image Process, 2017, 26(7): 3142-3155. doi: 10.1109/TIP.83
[15]	DOU Q, WANG Z, FENG X, et al. MRI denoising with a non-blind deep complex-valued convolutional neural network[J]. NMR Biomed, 2024, 38(1): e5291. doi: 10.1002/nbm.v38.1
[16]	JU R Y, CHEN C C, CHIANG J S, et al. Resolution enhancement processing on low quality images using swin transformer based on interval dense connection strategy[J]. Multimedia Tools Appl, 2023, 83(5): 14839-14855. doi: 10.1007/s11042-023-16088-0
[17]	CONG S, ZHOU Y. A review of convolutional neural network architectures and their optimizations[J]. Artif Intell Rev, 2022, 56(3):1905-1969. doi: 10.1007/s10462-022-10213-5
[18]	TIAN M, SONG K K. Boosting magnetic resonance image denoising with generative adversarial networks[J]. IEEE Access, 2021: 62266-62275.
[19]	JIANG D S, DOU W Q, VOSTERS L, et al. Denoising of 3D magnetic resonance images with multi-channel residual learning of convolutional neural network[J]. Jpn J Radiol, 2018, 36(9): 566-574. doi: 10.1007/s11604-018-0758-8 pmid: 29982919
[20]	NICHOL A Q, DHARIWAL P. Improved denoising diffusion probabilistic models[C]// MEILA M, ZHANG T. Proceedings of the 38th International Conference on Machine Learning. Cambridge, MA: PMLR, 2021, 139: 8162-8171.
[21]	CROITORU F A, HONDRU V, IONESCU R T, et al. Diffusion models in vision: a survey[J]. IEEE Trans Pattern Anal Mach Intell, 2023, 45(9): 10850-10869. doi: 10.1109/TPAMI.2023.3261988
[22]	LI S Y, XIONG H, CHEN Y Z. DiffCharge: Generating EV charging scenarios via a denoising diffusion model[J]. IEEE Trans Smart Grid, 2024, 15(4): 3936-3949. doi: 10.1109/TSG.2024.3360874
[23]	MA X, ZOU M, FANG X, et al. Convergent-diffusion denoising model for multi-scenario CT image reconstruction[J]. Comput Med Imaging Graph, 2025, 120: 102491. doi: 10.1016/j.compmedimag.2024.102491
[24]	KUANG G, KEITH J, GEORGES F E, et al. PET image denoising based on denoising diffusion probabilistic model[J]. Eur J Nucl Med Mol Imaging, 2023, 51(2): 358-368. doi: 10.1007/s00259-023-06417-8
[25]	CHEN X M, ZHAO X C, SUN X P, et al. Study on the automatic accumulation-thawing device of hyperpolarized ¹²⁹Xe[J]. Chinese J Magn Reson, 2022, 39(3): 316-326.
	陈小明, 赵修超, 孙献平, 等. 超极化 ¹²⁹Xe自动收集-升华装置研究[J]. 波谱学杂志, 2022, 39(3): 316-326. doi: 10.11938/cjmr20222998
[26]	CHEN Q, LI H D, FANG Y, et al. Association of ¹²⁹Xe ventilation functional MRI with pulmonary lesion types[J]. Chinese J Magn Reson, 2024, 41(3): 276-285.
	陈琪, 李海东, 方媛, 等. ¹²⁹Xe通气功能MRI与肺部病灶类型关联研究[J]. 波谱学杂志, 2024, 41(3): 276-285. doi: 10.11938/cjmr20243103
[27]	LI Z M, XIAO S, WANG C, et al. Complementation-reinforced network for integrated reconstruction and segmentation of pulmonary gas MRI with high acceleration[J]. Med Phys, 2023, 51(1): 378-393. doi: 10.1002/mp.16591 pmid: 37401205
[28]	ZHOU W, CONRAD A B, RAHIM H S, et al. Image quality assessment: from error visibility to structural similarity[J]. IEEE Trans Image Process, 2004, 13(4): 600-612. doi: 10.1109/TIP.2003.819861
[29]	LI Z M, XIAO S, WANG C, et al. Encoding enhanced complex CNN for accurate and highly accelerated MRI[J]. IEEE Trans Med Imaging, 2024, 43(5): 1828-1840. doi: 10.1109/TMI.2024.3351211
[30]	TAHIR A B, BRAGG M C, WILD M J, et al. Impact of field number and beam angle on functional image-guided lung cancer radiotherapy planning[J]. Phys Med Biol, 2017, 62(17): 7114-7130. doi: 10.1088/1361-6560/aa8074 pmid: 28800298
[31]	CHITWAN S, JONATHAN H, WILLIAM C, et al. Image super-resolution via iterative refinement[J]. IEEE Trans Pattern Anal Mach Intell, 2022, 45(4): 4713-4726.

方法	PSNR	SSIM
噪声图像	17.973±3.237**	0.512±0.163**
NLM	20.862±3.481**	0.584±0.079**
UNLM	24.319±3.125**	0.431±0.153**
DnCNN	30.758±2.739**	0.814±0.053**
HTC-net	31.915±2.452**	0.843±0.057**
DDPM	32.116±2.283*	0.827±0.055**
DDPM-HXe	33.264±2.594	0.847±0.058

方法	PSNR	SSIM
噪声图像	17.973±3.237**	0.512±0.163**
实验组一	32.032±2.174**	0.831±0.052**
实验组二	32.116±2.283**	0.827±0.055**
实验组三	32.694±2.272*	0.835±0.053**
实验组四	31.972±2.189**	0.819±0.052**
实验组五	33.264±2.594	0.847±0.058

超极化¹²⁹Xe MRI的多模态增强去噪扩散模型研究

Research on a Multi-modal Enhanced Denoising Diffusion Model for Hyperpolarized ¹²⁹Xe MRI

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 31

相关文章 15

编辑推荐

Metrics

本文评价

[1]	李英豪, 王丽辉, 王苏成, 朱中旗, 黄长栋, 李仁峰, 曹开明, 胡海洋, 贾一鸣, 梁松涛, 杨光, 路青, 汪红志. 胰腺自动分割与区域定量及糖尿病评估研究[J]. 波谱学杂志, 2025, 42(4): 378-389.
[2]	高照耀, 张展, 胡亮亮, 许光宇, 周胜, 胡雨欣, 林子捷, 周超. 基于虚拟线圈和GRAPPA增强网络的PMRI方法[J]. 波谱学杂志, 2025, 42(4): 390-401.
[3]	曹飞, 徐芊芊, 陈浩, 祖洁, 李晓文, 田锦, 鲍磊. 基于交叉自监督和DWI的NIID智能诊断方法[J]. 波谱学杂志, 2025, 42(2): 154-163.
[4]	薛培阳, 耿辰, 李郁欣, 鲍奕仿, 鲁宇澄, 戴亚康. 基于3D ResNet50改进模型的TOF-MRA脑动脉瘤分类方法[J]. 波谱学杂志, 2025, 42(1): 56-66.
[5]	李连杰, 郑瑜, 谢树光, 张鸣, 李红闯, 刘小玲, 赵修超, 韩叶清, 李海东, 范丽, 萧毅, 刘士远, 周欣. 超极化¹²⁹Xe MR用于COPD急性加重的肺功能研究[J]. 波谱学杂志, 2024, 41(4): 363-372.
[6]	宁欣宙, 黄臻, 陈西曲, 刘鑫杰, 陈罡, 张志, 鲍庆嘉, 刘朝阳. 用于超快时空编码MRI的Transformer超分辨率重建算法研究[J]. 波谱学杂志, 2024, 41(4): 454-468.
[7]	陈琪, 李海东, 方媛, 沈路阳, 刘无己, 罗明, 李业成, 张鸣, 赵修超, 石磊, 周倩, 韩叶清, 周欣. ¹²⁹Xe通气功能MRI与肺部病灶类型关联研究[J]. 波谱学杂志, 2024, 41(3): 276-285.
[8]	杨黎明, 王远军. 扩散张量图像去噪算法研究进展[J]. 波谱学杂志, 2024, 41(3): 341-361.
[9]	戴俊龙, 何聪, 武杰, 边云. 融合双解码和全局注意力上采样模块的胰腺囊性肿瘤分割网络[J]. 波谱学杂志, 2024, 41(2): 151-161.
[10]	杨钰, 陈博, 吴柳滨, 林恩平, 黄玉清, 陈忠. Laplace NMR谱图重建——从经典正则化到深度学习[J]. 波谱学杂志, 2024, 41(2): 191-208.
[11]	常博, 孙灏芸, 高清宇, 王丽嘉. 传统方法和深度学习用于不同模态心脏医学图像的分割研究进展[J]. 波谱学杂志, 2024, 41(2): 224-244.
[12]	徐真顺, 袁小涵, 黄子珩, 邵成伟, 武杰, 边云. 基于深度学习的胰腺黏液性和浆液性囊性肿瘤的多源特征分类模型[J]. 波谱学杂志, 2024, 41(1): 19-29.
[13]	周敏雄, 戚轩, 杜兵, 齐东, 王海杰, 杨光, 蔡文梅, 刘孟潇, 张会婷, 严序, 聂生东, 何永胜. b值范围对6种体部扩散模型在前列腺应用的影响评估[J]. 波谱学杂志, 2024, 41(1): 9-18.
[14]	赖嘉雯, 汪宇玲, 蔡晓宇, 周丽华. 基于CNN-SVM的多维度信息融合半月板撕裂分类方法[J]. 波谱学杂志, 2023, 40(4): 423-434.
[15]	王慧, 王甜甜, 王丽嘉. 基于心脏磁共振电影图像的压缩激励残差U形网络左心肌分割[J]. 波谱学杂志, 2023, 40(4): 435-447.