ISMRM25 - Diffusion Models Across the MRI Spectrum

ISMRM & ISMRT Annual Meeting & Exhibition • 10-15 May 2025 • Honolulu, Hawai'i

Diffusion Models Across the MRI Spectrum

Digital Poster

AI & Machine Learning

Wednesday, 14 May 2025

Exhibition Hall

16:45 - 17:45

Session Number: D-24

No CME/CE Credit

		Computer Number: 1 3971. Unconditional Diffusion Model for 3D MRI Artifact Removal and Detail Refinement H. Deng, Z. Zhu, L. Zhang, M. Yang, Z. Lan, M. Yang, S. Wang, J. Liu, H. Zhang School of Biomedical Engineering , ShanghaiTech University, Shanghai, 201210, China, Shanghai, China Impact: We propose a motion artifact reduction method based on unconditional DPM with a supervised fine-tuning module, DR. This approach demonstrates significant accuracy and robustness. Our method is highly valuable to neuroscience and clinical studies on existing and future large-scale datasets.
		Computer Number: 3972. WITHDRAWN
		Computer Number: 2 3973. Universal Denoising for MRI DICOM Images Across Diverse Clinical Conditions Through Variational Diffusion Model View Presentation Video Y. Shao, Y. Qiu, D. Li, L. Zhang, S. Lin, S. Chen, P. Chen, X. Jin, L. Du, Y. Gu, X. Huang, A. Liu, J. Zhong, S. Liao, K. Sun, D. Shen ShanghaiTech University, Shanghai, China Impact: This study provides a novel solution to the over-smoothness issue for diffusion models when dealing with diverse and complex real-world data. Our model shows promising denoising performance on real-world clinical images scanned with 2x or 3x acceleration factor.
		Computer Number: 3 3974. Unsupervised Zero-Shot MR Image Denoising via Diffusion Null-space Model K. Zhang, A. Shankaranarayanan, Z. Zhou University of Washington, Seattle, United States Impact: The proposed zero-shot method enables immediate deployment across different MRI protocols without retraining, enabling reduced scan times or lower field strengths while maintaining diagnostic quality. The approach opens possibilities for applying diffusion models to MRI challenges without extensive data collection.
		Computer Number: 4 3975. High-speed high-quality super-resolution diffusion model via flow-matching View Presentation Video J. Park, J. Yoon, M. Kim, J. Choi, J. Lee Seoul National University, Seoul, Korea, Republic of Impact: The proposed data-consistency guided super-resolution network using diffusion model via flow matching demonstrates high-quality images, outperforming existing methods while shortening the processing times more than 10 times (43.9 sec vs. 8 min 4 sec).
		Computer Number: 5 3976. Motion correction and scan acceleration using diffusion informed prior View Presentation Video G. Son, D. Kim Yonsei university, Seoul, Korea, Republic of Impact: This framework provides robust motion correction for MRI, preserving details and enabling broader applications for complex-domain MRI data.
		Computer Number: 6 3977. Graph-Aided Diffusion Models for Precision Segmentation of Challenging Tumor Areas in MRI View Presentation Video L. Rivera Monroy, L. Pfaff, T. Wang, N. Kleuser, A. Maier Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Erlangen, Germany Impact: This study builds on prior work by combining diffusion models with graph neural networks to enhance brain tumor segmentation. By utilizing focused graph representations, the proposed method improves precision particularly within the tumor core and smaller subregions.
		Computer Number: 7 3978. Diffusion Probabilistic Generative Models for Accelerated in-NICU, Permanent Magnet Neonatal MRI Reconstruction View Presentation Video Y. Arefeen, B. Levac, J. Tamir The University of Texas at Austin, Austin, United States Impact: The improvement in acquisition speed of T₁ and T₂ weighted lower field neonatal MRI protocols using diffusion-probabilistic-generative models, trained with methods designed to handle the noisy, limited data, improves accessibility of MRI to patients in the neonatal-intensive-care-unit.
		Computer Number: 8 3979. Blind harmonization reinvented via 3D diffusion model View Presentation Video H. Jeong, H. Lee, S. Y. Chun, J. Lee Seoul National University, Seoul, Korea, Republic of Impact: BlindHarmonyDiff is a novel blind harmonization technique that overcomes the limitations of previous methods by generating high-quality 3D harmonized images, effectively handling images with large domain gap. Its refinement module mitigates hallucinations from diffusion models, improving reliability and clinical applicability.
		Computer Number: 9 3980. Correction of Motion-Affected MRI Images via Motion-Adaptive Diffusion Model View Presentation Video Q. Sun, Y. Chai, M. Khoo, H. Kim Neuroimaging with Deep Learning Lab (NIDLL), Stevens Neuroimaging and Informatics Institute, University of Southern California, Los Angeles , United States Impact: This project significantly improves MRI diagnostic accuracy by effectively correcting motion artifacts. It provides a more efficient and reliable solution for both clinical and research applications by reducing the need for repeat scans and enhancing image quality.
		Computer Number: 10 3981. Efficient and Effective Control Adapters for Diffusion Plug-and-Play models: A DWI Application View Presentation Video B. Xin, R. Miron, M. Mostapha, N. Janardhanan, O. Darwish, T. Huelnhagen, T. Wuerfl, D. Grodzki, R. Schneider, M. Nadar Rutgers, The State University of New Jersey, NJ, USA, New Jersey, United States Impact: The proposed PnP method offers a flexible solution for using a pre-trained diffusion prior in a flexible framework for image reconstruction. Using the adapter can improve certain scenarios, only if needed, without the need for re-training the prior.
		Computer Number: 11 3982. VMamba UNet with Latent Diffusion Prior Guidance for MRI Reconstruction View Presentation Video L. Zhang, B. Qiu University of Science and Technology of China, Hefei, China Impact: This work investigates the role of generative priors of latent diffusion models on MRI reconstruction tasks and facilitates the frontier studies of deep-learning-based rapid reconstruction methods.
		Computer Number: 12 3983. Cine Cardiac MR Super-resolution using a Fast Diffusion Model with Motion Guided Temporal Consistency Enforcement View Presentation Video H. Liao, C. Liu, Q. Miao, P. Hu, H. Qi ShanghaiTech University, Shanghai, China Impact: The proposed method yielding good-quality cine cardiac MR image series from low-resolution images enables accelerated cine cardiac MR acquisition, and could be potentially applied to achieve high spatial-temporal real-time cardiac MRI.
		Computer Number: 13 3984. No Annotate Again (NAA): Realistic Image and Annotation Synthesis for Multi-Contrast MRI through Diffusion without Paired Data PDF Poster View Presentation Video X. Chen, C. Li, E. Chen, Y. Liu, L. Zhao, T. Chen, S. Sun UII America Inc., Boston, United States Impact: NAA enables scalable, annotation-free neural network developments for medical image analysis. This approach reduces dependency on annotated datasets and can benefit a wide range of applications.
		Computer Number: 14 3985. Contrast-enhanced Cardiac MRI Generation Using Contrast-free Cardiac Cine MRI P. Xie, Z. Li, W. Chen, Y. Ma, J. Xiao Chongqing Emergency Medical Centre, Chongqing University Central Hospital, School of Medicine, Chongqing University, Chongqing, China Impact: We demonstrate the potential of diffusion models in generating LGE using contrast-free cine images. The proposed method is expected to provide a valuable alternative for patients who cannot obtain LGE images due to contraindications.
		Computer Number: 15 3986. A Unified Diffusion Model for Multimodal Image Reconstruction and Synthesis View Presentation Video W. Gan, X. Wang, T. Wang, C. Ying, Y. Hu, Y. Chen, H. An, U. Kamilov Washington University in St. Louis, St. Louis, United States Impact: This work simplifies multimodality medical imaging by using one model for reconstruction and synthesis, reducing training and deployment complexity. It also offers the potential to enhance clinical imaging efficiency and provides a new tool for leveraging multimodality information.

Similar Session(s)