ISMRM25 - Transformative Diffusion Models for MRI

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

Transformative Diffusion Models for MRI

Oral

AI & Machine Learning

Wednesday, 14 May 2025

323AB

13:30 - 15:30

Moderators: Jonathan Tamir, Martin Uecker & Lei Qiu

Session Number: O-10

No CME/CE Credit

13:30		Introduction
13:42		0928. Face Decoding and Reconstruction from 7T Laminar fMRI Data using A Diffusion Generative Model View Presentation Video N. Huynh, G. Deshpande Auburn University, Auburn, United States Impact: Brain disorders, like stroke and prosopagnosia, can impair brain regions for facial processing, making face perception difficult. By understanding the neural circuitry involved in face perception, researchers may identify pathways that could be targeted to alleviate symptoms in these conditions
13:54		0929. Diffusion based multi-domain neuroimaging harmonization method with preservation of anatomical details View Presentation Video H. Lan, B. Varghese, N. Sheikh-Bahaei, F. Sepehrband, A. Toga, J. Choupan University of Southern California, Los Angeles, United States Impact: This work showed efficacy of using diffusion model to tackle neuroimaging harmonization problem with the preservation of anatomical and biological details. It is specifically evaluated to harmonize the imaging texture heterogeneity present in the large cohorts of multi-center dataset.
14:06		0930. Multidimensional MR Image Reconstruction Using A Disentangled Representation View Presentation Video R. Zhao, F. Lam University of illinois, Urbana Chamapign, Champaign, United States Impact: The proposed method may provide a new perspective for learning-based, high-dimensional MRI reconstruction, for which small or even no data are available problem-specific supervised training.
14:18		0931. Enhancing Pathological Fetal MRI Segmentation through Generative AI: A Novel Approach to Synthetic Pathological Data Generation View Presentation Video M. Kaandorp, H. Asma-ull, H-G Kim, D. Agbelese, K. Payette, A. Jakab University Children’s Hospital Zurich, Zurich, Switzerland Impact: Our approach overcomes challenges of limited annotated pathological MRI datasets, facilitating the training of robust segmentation models without the need for pathological data. This advancement is an important step towards addressing privacy issues while improving segmentation performance in prenatal imaging.
14:30		0932. LGEDiffusion: A Multi-Sequence Guided Diffusion Model for Virtual Contrast-Free LGE Generation in Myocardial Infarction View Presentation Video J. Qi, X. Yue, M. Hu, J. Li, T. Li, K. He Medical Innovation Research Department, Chinese PLA General Hospital, Beijing, China Impact: This study highlights the potential of denoising diffusion probabilistic models for multi-sequence-guided MRI translation, emphasizing the value of virtual LGE as a viable contrast-free imaging alternative for myocardial infarction assessment.
14:42		0933. A Self-Consistent Diffusion Schrodinger Bridge for Multi-Modal Medical Image Translation View Presentation Video F. Arslan, B. Kabas, O. Dalmaz, M. Ozbey, T. Cukur Bilkent University, Ankara, Turkey Impact: The enhanced image fidelity in multi-modal protocols achieved by SelfRDB can extend the scope of imaging-based assessments, while maintaining relatively low scan budgets and minimizing exposure to invasive agents or radiation, particularly benefiting at-risk pediatric and elderly populations.
14:54		0934. Accelerating Longitudinal MRI using Prior Informed Latent Posterior Sampling (PIPS) View Presentation Video Z. Shah, Y. Urman, A. Kumar, B. Soares, K. Setsompop Stanford University, Stanford, United States Impact: We propose an unsupervised prior conditioning method to further accelerate MRI for longitudinal studies. Our method is both scalable and generalizable, as it does not require sequential k-space for training and enforces data consistency throughout the reconstruction.
15:06		0935. Variational Diffusion Models for Motion Correction: Comprehensive Evaluation View Presentation Video J. Oscanoa, C. Alkan, A. Nurdinova, D. Abraham, K. Setsompop, M. Mardani, D. Ennis, J. Pauly, S. Vasanawala Stanford University, Stanford, United States Impact: We demonstrate the value of our blind inverse problem framework based on diffusion models. Our method outperforms state-of-the-art methods for reconstruction with motion correction in both retrospectively and prospectively corrupted data.
15:18		0936. AI-powered 0.3 mm Ultrahigh Resolution MR Brain Imaging View Presentation Video Z. Ke, Z. Xu, H. Zhuang, W. Tang, Y. Li, Z-P Liang University of Illinois at Urbana-Champaign, Urbana, United States Impact: Conventional MRI scans of the brain are typically done at 1 mm resolution. Ultrahigh-resolution MRI will open up many opportunities for research and clinical applications. The proposed approach may also be useful for solving other imaging and processing problems.

Similar Session(s)