Cape Town - 2026 ISMRM-ISMRT Annual Meeting and Exhibition - Novel Reconstruction Techniques for Fast Imaging

Novel Reconstruction Techniques for Fast Imaging

Oral

Acquisition & Reconstruction

Wednesday, 13 May 2026

Meeting Room 1.60

08:20 - 10:10

Moderators: Prakash Kumar & Carlos Castillo-Passi

Session Number: 508-02

CME/CE Credit Available

Similar Sessions

This session covers fast imaging techniques, including real-time, dynamic, and accelerated imaging, with an emphasis on novel reconstruction algorithms that enable high-speed data acquisition and image formation.

Skill Level: Intermediate,Advanced

08:20		508-02-001. Memory-Efficient Iterative Subspace Reconstructions on GPUs for Non-Cartesian MRI Ivo Maatman, Moritz Blumenthal, Nick Scholand, Sebastian Flassbeck, Martin Uecker, Jakob Assländer Center for Biomedical Imaging, Department of Radiology, NYU Grossman School of Medicine, New York, United States of America Impact: We developed three methods to reduce total memory requirements of Toeplitz-embedded subspace reconstructions by a factor of approximately 8; enabling high-resolution iterative reconstructions and machine learning on GPUs. The code is freely available in BART and Julia.
08:31		508-02-002. Advancing EPTI by rapid sequence prototyping, adaptive field correction, and scalable cloud-powered real-time reconstruction Jian Wu, Timothy Reese, Michael Hansen, Bruce Rosen, Lawrence Wald, Zijing Dong, Fuyixue Wang Massachusetts General Hospital and Harvard Medical School, Boston, United States of America Impact: We developed EPTI on vendor-neutral Pulseq supported by a scalable cloud-based real-time reconstruction framework for its broader clinical/neuroscientific adoption. We demonstrated rapid prototyping of Pulseq-EPTI in qMRI, fMRI, and CEST, and the cloud-based computation framework enables near-real-time EPTI reconstruction.
08:42		508-02-003. One-Heartbeat Cardiac Cine MRI via Phase-Guided Conditional Diffusion Hanrui Shi, Jian Xu, Qi Liu, Hongyu Li University of Washington, Seattle, United States of America Impact: This study enables one-heartbeat cardiac cine MRI at real-time acceleration by integrating phase-guided conditional diffusion. It offers faster acquisition with reduced motion and breath-hold dependence, improving clinical efficiency and patient comfort.
08:53		508-02-004. 1-minute free-breathing abdominal T1-weighted MRI in clinical practice using deep-learning auto-navigation and reconstruction Victor Murray, Yan Wen, Subin Erattakulangara, Oguz Akin, Richard Do, Gerald Behr, Arnaud Guidon, Ricardo Otazo Memorial Sloan Kettering Cancer Center, New York, United States of America Impact: The Movienet clinical prototype enables 1-minute free-breathing abdominal T1-weighted MRI with superior performance compared to clinical product technology, as confirmed by expert radiologists. Movienet is promising to improve efficiency and overall quality in regions affected by motion.
09:04		508-02-005. End-to-End Framework for Real-Time Image Reconstruction, Device and Tissue Tracking in MRI-Guided Liver Interventions Wenqi Zhou, Qing Dai, Christina Kerr, Shu-Fu Shih, Timoteo Delgado, Tsu-Chin Tsao, David Lu, Jason Chiang, Holden Wu David Geffen School of Medicine, University of California Los Angeles, Los Angeles, United States of America Impact: An end-to-end framework integrating real-time MRI reconstruction, needle tracking, and tissue tracking was established. The framework achieved accurate tracking results with end-to-end latency under 200 ms/frame, demonstrating potential to assist real-time MRI-guided liver interventions.
09:15		508-02-006. An Accelerated Deep Image Prior Reconstruction for Cardiac MR Fingerprinting Using Meta-Learning Zhongnan Liu, Calder Sheagren, Zexuan Liu, Nicole Seiberlich, Liyue Shen, Jesse Hamilton University of Michigan, Ann Arbor, United States of America Impact: A strategy is proposed to accelerate a zero-shot deep learning (Deep Image Prior) reconstruction for cardiac MRF using meta-learning followed by scan-specific fine-tuning, reducing the reconstruction time from 40 to 2.4 minutes while maintaining quantitative T₁ and T₂ mapping accuracy.
09:26		508-02-007. Radial ML-DIP: Real-Time 3D Cine Cardiovascular MRI Using Multi-Dynamic Low-Rank Deep Image Priors with Radial Sampling Jianli Wei, Xavier Sieber, Jérôme Yerly, YINGMIN LIU, Yixuan Liu, Katherine Binzel, Juliet Varghese, Ruud van Heeswijk, Matthias Stuber, Orlando Simonetti, Rizwan Ahmad, Samuel Ting The Ohio State University, Columbus, United States of America Impact: Real-time 3D cine MRI using radial ML-DIP provides isotropic spatial resolution and dense sampling of center k-space region for improved temporal regularization. We achieve free-breathing 3D cine reconstruction at 0.55T with 0.98 mm3 spatial and 82.8 ms temporal resolution.
09:37		508-02-008. Joint Implicit Neural Representation for Fast Scan-Specific Magnetic Resonance Fingerprinting Hongze Yu, Christopher Keen, Kaixuan Jin, Jeffrey Fessler, Yun Jiang University of Michigan, Ann Arbor, United States of America Impact: The method moves toward clinically practical MRF by reducing acquisition and reconstruction time without sacrificing T1/T2 accuracy or precision. By scan specifically learning a shared anatomical prior, it enables fast quantitative mapping and may extend to other quantitative MRI protocols.
09:48		508-02-009. Open-Source Pulseq Reduced-FOV Diffusion MRI for Quantitative Mapping of Unilateral Lumbosacral Radiculopathy Evgenios Kornaropoulos, Pierre Pesesse, Mark Vanderthommen, Christophe Demoulin, Lamalle Laurent, Christophe Phillips, Mikhail Zubkov GIGA-Institute, University of Liège, Liège, Belgium Impact: This open-source Pulseq framework, unlike vendor-based DTI, exposes full control to optimize SNR and minimize distortion-induced blurring, delivering reproducible reduced-FOV diffusion of lumbosacral roots with improved sensitivity to unilateral radiculopathy across scanners and sites.
09:59		508-02-010. Guided Discussion of Fast Imaging Reconstruction Techniques Prakash Kumar University of Southern California, Los Angeles, United States of America