Self‐supervised adversarial diffusion models for fast MRI reconstruction
Medical Physics · 2025
DOI: 10.1002/mp.17675 ↗At a glance
Journal article in Medical Physics, 2025. Written by Mojtaba Safari and 4 co-authors. Institution: Department of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, Georgia, USA. Major MEDLINE topics: Image Processing, Computer-Assisted, Magnetic Resonance Imaging, Deep Learning. Cited 10 times so far (Crossref); 44 references. Funded by: NIH HHS, National Institutes of Health.
Abstract (original language)
Background
Magnetic resonance imaging (MRI) offers excellent soft tissue contrast essential for diagnosis and treatment, but its long acquisition times can cause patient discomfort and motion artifacts.
Purpose
To propose a self‐supervised deep learning‐based compressed sensing MRI method named “Self‐Supervised Adversarial Diffusion for MRI Accelerated Reconstruction (SSAD‐MRI)” to accelerate data acquisition without requiring fully sampled datasets. Materials and Methods We used the fastMRI multi‐coil brain axial ‐weighted (‐w) dataset from 1376 cases and single‐coil brain quantitative magnetization prepared 2 rapid acquisition gradient echoes maps from 318 cases to train and test our model. Robustness against domain shift was evaluated using two out‐of‐distribution (OOD) datasets: multi‐coil brain axial postcontrast ‐weighted () dataset from 50 cases and axial T1‐weighted (T1‐w) dataset from 50 patients. Data were retrospectively subsampled at acceleration rates . SSAD‐MRI partitions a random sampling pattern into two disjoint sets, ensuring data consistency during training. We compared our method with ReconFormer Transformer and SS‐MRI, assessing performance using normalized mean squared error (NMSE), peak signal‐to‐noise ratio (PSNR), and structural similarity index (SSIM). Statistical tests included one‐way analysis of variance and multi‐comparison Tukey's honesty significant difference (HSD) tests.
Results
SSAD‐MRI preserved fine structures and brain abnormalities visually better than comparative methods at for both multi‐coil and single‐coil datasets. It achieved the lowest NMSE at , and the highest PSNR and SSIM values at all acceleration rates for the multi‐coil dataset. Similar trends were observed for the single‐coil dataset, though SSIM values were comparable to ReconFormer at . These results were further confirmed by the voxel‐wise correlation scatter plots. OOD results showed significant ( p ) improvements in undersampled image quality after reconstruction.
Conclusions
SSAD‐MRI successfully reconstructs fully sampled images without utilizing them in the training step, potentially reducing imaging costs and enhancing image quality crucial for diagnosis and treatment.
Subjects
MeSH terms (MEDLINE)
Author keywords
Authors and institutions
- Department of Radiation Oncology and Winship Cancer Institute, Emory University, Atlanta, Georgia, USA.
Funding
- NIH HHS
- National Institutes of Health
Cite
APA
Safari, M., Eidex, Z., Pan, S., Qiu, R. L. J., & Yang, X. (2025). Self‐supervised adversarial diffusion models for fast MRI reconstruction. Medical Physics. https://doi.org/10.1002/mp.17675
Vancouver
Safari M, Eidex Z, Pan S, Qiu RLJ, Yang X. Self‐supervised adversarial diffusion models for fast MRI reconstruction. Medical Physics. 2025. doi:10.1002/mp.17675
BibTeX
@article{Safari2025,
author = {Safari, Mojtaba and Eidex, Zach and Pan, Shaoyan and Qiu, Richard L J and Yang, Xiaofeng},
title = {Self‐supervised adversarial diffusion models for fast MRI reconstruction},
journal = {Medical Physics},
year = {2025},
doi = {10.1002/mp.17675}
}
EUVIMED