Objective To develop a method for generating 7T susceptibility weighted image (SWI) from 3T SWI data for more accurate vascular structure reconstruction. Methods We propose a vessel-enhanced consistency-preserving diffusion model for synthesizing 7T SWI. This method first utilizes a pre-trained vessel segmentation network to extract vessel maps from both 3T and 7T SWI as vascular priors. The vessel maps from 7T SWI are employed to construct an intra-slice vessel-enhanced loss and reinforce the reconstructed fine vascular structures. An inter-slice feature fusion module and a vessel-aware consistency constraint are integrated into the diffusion model to enhance the spatial coherence of the generated results. The feature fusion module integrates contextual features from adjacent slices via a cross-attention mechanism, enabling deep information interaction across multiple slices. The vessel-aware consistency constraint employs an inter-slice vessel-aware consistency-preserving loss to impose constraints on overlapping regions of the sliding windows, thereby enhancing generation coherence across slices. Results The experiment conducted on 209 paired 3T-7T SWI datasets showed that, compared to the suboptimal method EDM, the proposed method increased the PSNR by 1.39 dB, improved the SSIM by 2.28%, enhanced the skeleton coverage similarity (SCS) by 6.57%, and reduced MAE and LPIPS by 15.84% and 17.07%, respectively. Qualitative analysis further validated the model's superiority in generating vascular details, including accurate restoration of major venous structures and clear delineation of low-contrast microvessels, with natural continuous vascular branches. Ablation studies confirmed the effectiveness of each individual module and their contribution to the performance gains. Conclusion The proposed method can effectively synthesize 7T SWI images with clear vascular details from 3T SWI to facilitate clinical and scientific analyses while balancing computational resource requirements and image reconstruction quality.
BarisanoG, SepehrbandF, MaS, et al. Clinical 7 T MRI: Are we there yet? A review about magnetic resonance imaging at ultra-high field[J]. Br J Radiol, 2019, 92(1094): 20180492. doi:10.1259/bjr.20180492
[6]
LiuSF, BuchS, ChenYS, et al. Susceptibility-weighted imaging: current status and future directions[J]. NMR Biomed, 2017, 30(4): e3552. doi:10.1002/nbm.3552
[7]
ForstmannBU, IsaacsBR, TemelY. Ultra high field MRI-guided deep brain stimulation[J]. Trends Biotechnol, 2017, 35(10): 904-7. doi:10.1016/j.tibtech.2017.06.010
LaddME. High-field-strength magnetic resonance: potential and limits[J]. Top Magn Reson Imag, 2007, 18(2): 139-52. doi:10.1097/rmr.0b013e3180f612b3
[10]
ZhangD, DuanCH, AnazodoU, et al. Self-supervised anatomical continuity enhancement network for 7T SWI synthesis from 3T SWI[J]. Med Image Anal, 2024, 95: 103184. doi:10.1016/j.media.2024.103184
[11]
BahramiK, ShiF, ZongXP, et al. Hierarchical reconstruction of 7T-like images from 3T MRI using multi-level CCA and group sparsity[C]//Medical Image Computing and Computer-Assisted Intervention -- MICCAI 2015. Cham: Springer, 2015: 659-66. doi:10.1007/978-3-319-24571-3_79
[12]
ZhangYQ, ChengJZ, XiangL, et al. Dual-domain cascaded regression for synthesizing 7T from 3T MRI[C]//Medical Image Computing and Computer Assisted Intervention-MICCAI 2018. Cham: Springer, 2018: 410-7. doi:10.1007/978-3-030-00928-1_47
[13]
BahramiK, ShiF, RekikI, et al. Convolutional neural network for reconstruction of 7T-like images from 3T MRI using appearance and anatomical features[C]//Deep Learning and Data Labeling for Medical Applications. Cham: Springer, 2016: 39-47. doi:10.1007/978-3-319-46976-8_5
[14]
LiuXL, QiuS, ZhouM, et al. WFANet-DDCL: wavelet-based frequency attention network and dual domain consistency learning for 7T MRI synthesis from 3T MRI[J]. IEEE Trans Circuits Syst Video Technol, 2025, 35(6): 5617-32. doi:10.1109/tcsvt.2025.3536807
[15]
LiS, DengXG, LiQW, et al. Synthetizing SWI from 3T to 7T by generative diffusion network for deep medullary veins visualization[J]. NeuroImage, 2025, 320: 121475. doi:10.1016/j.neuroimage.2025.121475
[16]
NieD, TrulloR, LianJ, et al. Medical image synthesis with deep convolutional adversarial networks[J]. IEEE Trans Biomed Eng, 2018, 65(12): 2720-30. doi:10.1109/tbme.2018.2814538
[17]
DuanCH, BianXB, ChengK, et al. Synthesized 7T MPRAGE from 3T MPRAGE using generative adversarial network and validation in clinical brain imaging: a feasibility study[J]. Magnetic Resonance Imaging, 2024, 59(5): 1620-9. doi:10.1002/jmri.28944
[18]
WeiJ, PanYS, XiaY, et al. Learning to synthesize 7 T MRI from 3 T MRI with few data by deformable augmentation[C]//Machine Learning in Medical Imaging. Cham: Springer, 2021: 70-9. doi:10.1007/978-3-030-87589-3_8
[19]
TrattnigS, BognerW, GruberS, et al. Clinical applications at ultrahigh field (7 T). Where does it make the difference[J]? NMR Biomed, 2016, 29(9): 1316-34. doi:10.1002/nbm.3272
[20]
LiuCL, LiW, TongKA, et al. Susceptibility-weighted imaging and quantitative susceptibility mapping in the brain: Brain Susceptibility Imaging and Mapping[J]. J Magn Reson Imaging, 2015, 42(1): 23-41. doi:10.1002/jmri.24768
[21]
BillotB, GreveDN, PuontiO, et al. SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining[J]. Med Image Anal, 2023, 86: 102789. doi:10.1016/j.media.2023.102789
[22]
AvantsBB, TustisonNJ, SongG, et al. A reproducible evaluation of ANTs similarity metric performance in brain image registration[J]. NeuroImage, 2011, 54(3): 2033-44. doi:10.1016/j.neuroimage.2010.09.025
KawarB, EladM, ErmonS, et al. Denoising diffusion restoration models[J]. Adv Neural Inf Process Syst, 2022, 35: 23593-606. doi:10.52202/068431-1714
[25]
CaoHQ, TanC, GaoZY, et al. A survey on generative diffusion models[J]. IEEE Trans Knowl Data Eng, 2024, 36(7): 2814-30. doi:10.1109/tkde.2024.3361474
[26]
HatamizadehA, NathV, TangYC, et al. Swin UNETR: swin transformers for semantic segmentation of brain tumors in MRI images[M]//Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries. Cham: Springer International Publishing, 2022: 272-84. doi:10.1007/978-3-031-08999-2_22
[27]
KarrasT, AittalaM, AilaT, et al. Elucidating the design space of diffusion-based generative models[J]. Adv Neural Inf Process Syst, 2022, 35: 26565-77. doi:10.52202/068431-1926
[28]
QuLQ, ZhangYQ, WangS, et al. Synthesized 7T MRI from 3T MRI via deep learning in spatial and wavelet domains[J]. Med Image Anal, 2020, 62: 101663. doi:10.1016/j.media.2020.101663
ZhuJY, ParkT, IsolaP, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks[C]//2017 IEEE International Conference on Computer Vision (ICCV). October 22-29, 2017. Venice. IEEE, 2017: 2223-32. doi:10.1109/iccv.2017.244
[31]
RutlandJW, DelmanBN, GillCM, et al. Emerging use of ultra-high-field 7T MRI in the study of intracranial vascularity: state of the field and future directions[J]. AJNR Am JNeuroradiol, 2020, 41(1): 2-9. doi:10.3174/AJNR.A6344
[32]
YangX, LiuCY, Le MinhH, et al. An automated method for accurate vessel segmentation[J]. Phys Med Biol, 2017, 62(9): 3757. doi:10.1088/1361-6560/aa6418