Magnetic Resonance Imaging Acceleration Using Diffusion Models::  A Comparative Study of MC-DDPM and HFS-SDE

João Paulo Ferreira Junior; Renata Manzano Maria; Marcelo Andrade da Costa Vieira

doi:10.22456/2175-2745.150968

Authors

João Paulo Ferreira Junior Universidade de São Paulo (USP) https://orcid.org/0009-0007-3952-5882
Renata Manzano Maria Universidade de São Paulo (USP) https://orcid.org/0000-0002-2119-8269
Marcelo Andrade da Costa Vieira Universidade de São Paulo (USP) https://orcid.org/0000-0002-6038-7740

DOI:

https://doi.org/10.22456/2175-2745.150968

Keywords:

magnetic resonance imaging, image reconstruction, diffusion models, deep learning

Abstract

Magnetic resonance imaging (MRI) reconstruction from undersampled data is central to scan acceleration. This study compares the diffusion models MC-DDPM and HFS-SDE for MRI reconstruction, trained and tested on the same dataset under identical experimental conditions, thereby addressing a common heterogeneity in the literature, where results are reported with different datasets, sampling masks, and input formats, hindering direct comparison. Acceleration factors R = 4, 8, 12, and 16 are evaluated using objective image-quality metrics (PSNR, SSIM, and NMSE). Results show that MC-DDPM consistently outperforms HFS-SDE across all scenarios, with larger gaps at higher accelerations. These findings indicate that standardized protocols make the relative strengths and limitations of the evaluated diffusion-based models clearer.

Downloads

Download data is not yet available.

References

[1] ZAITSEV, M.; MACLAREN, J.; HERBST, M. Motion artifacts in MRI: A complex problem with many partial solutions. Journal of Magnetic Resonance Imaging, Wiley, v. 42, n. 4, p. 887–901, 2015.

[2] GRISWOLD, M. A. et al. Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magnetic Resonance in Medicine, v. 47, n. 6, p. 1202–1210, 2002.

[3] LUSTIG, M.; DONOHO, D.; PAULY, J. M. Sparse MRI: The application of compressed sensing for rapid MR imaging. Magnetic Resonance in Medicine, Wiley Online Library, v. 58, n. 6, p. 1182–1195, 2007.

[4] HAMMERNIK, K. et al. Learning a variational network for reconstruction of accelerated MRI data. Magnetic Resonance in Medicine, Wiley Online Library, v. 79, n. 6, p. 3055–3071, 2018.

[5] SCHLEMPER, J. et al. A deep cascade of convolutional neural networks for dynamic MR image reconstruction. IEEE Transactions on Medical Imaging, IEEE, v. 37, n. 2, p. 491–503, 2018.

[6] MUCKLEY, M. J. et al. Results of the 2020 fastMRI challenge for machine learning MR image reconstruction. IEEE Transactions on Medical Imaging, IEEE, v. 40, n. 9, p. 2306–2317, 2021.

[7] AGGARWAL, H. K.; MANI, M. P.; JACOB, M. MoDL: Model-based deep learning architecture for inverse problems. IEEE Transactions on Medical Imaging, v. 38, n. 2, p. 394–405, 2019. Disponível em: https://pubmed.ncbi.nlm.nih.gov/30106719/.

[8] SRIRAM, A. et al. End-to-end variational networks for accelerated MRI reconstruction. NeurIPS (Datasets and Benchmarks Track) / arXiv, 2020. Disponível em: https://arxiv.org/abs/2004.06688.

[9] ZBONTAR, J. et al. fastMRI: An open dataset and benchmarks for accelerated MRI. arXiv preprint arXiv:1811.08839, 2018. Disponível em: https://arxiv.org/abs/1811.08839.

[10] HO, J.; JAIN, A.; ABBEEL, P. Denoising diffusion probabilistic models. In: Advances in Neural Information Processing Systems (NeurIPS). [s.n.], 2020. v. 33, p. 6840–6851. Disponível em: https://proceedings.neurips.cc/paper_files/paper/2020/hash/4c5bcfec8584af0d967f1ab10179ca4b-Abstract.html.

[11] SONG, Y. et al. Score-based generative modeling through stochastic differential equations. In: International Conference on Learning Representations (ICLR). [s.n.], 2021. Disponível em: https://openreview.net/forum?id=PxTIG12RRHS.

[12] XIE, Y.; LI, Q. Measurement-conditioned denoising diffusion probabilistic model for under-sampled medical image reconstruction. arXiv preprint arXiv:2203.03623, 2022. Disponível em: https://arxiv.org/abs/2203.03623.

[13] CAO, C. et al. High-frequency space diffusion model for accelerated MRI. IEEE Transactions on Medical Imaging, 2024. Disponível em: https://pubmed.ncbi.nlm.nih.gov/38194398/.

[14] XIANG, T. et al. DiffCMR: Fast cardiac MRI reconstruction with diffusion probabilistic models. arXiv preprint arXiv:2312.04853, 2023. Disponível em: https://arxiv.org/abs/2312.04853.

[15] PARK, C. Y. et al. Measurement score-based diffusion model. arXiv preprint arXiv:2505.11853, 2025. Disponível em: https://arxiv.org/abs/2505.11853.

[16] ZBONTAR, J.; KNOLL, F. et al. fastMRI Dataset. https://fastmri.med.nyu.edu/. Acesso em: 26 de fevereiro de 2026.

[17] HORE, A.; ZIOU, D. Image quality metrics: PSNR vs. SSIM. In: 2010 20th International Conference on Pattern Recognition. [S.l.: s.n.], 2010. p. 2366–2369.

[18] WANG, Z. et al. Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Processing, v. 13, n. 4, p. 600–612, 2004.

[19] DOHMEN, M. et al. Similarity and quality metrics for MR image-to-image problems. Scientific Reports, v. 15, n. 1, p. 87358, 2025. Disponível em: https://www.nature.com/articles/s41598-025-87358-0.