Cape Town - 2026 ISMRM-ISMRT Annual Meeting and Exhibition - AI: Emerging Techniques and Clinical Applications

AI: Emerging Techniques and Clinical Applications

Oral

Acquisition & Reconstruction

Wednesday, 13 May 2026

Auditorium 1

16:00 - 17:50

Moderators: Christoph Kolbitsch & Rebecca Baker

Session Number: 503-03

No CME/CE Credit

Similar Sessions

This session highlights emerging AI techniques in MRI and their clinical applications, focusing on image reconstruction and accelerated acquisition.

16:00		503-03-001. All-in-One DeepGrasp: A Unified Self-Supervised Model for Accelerated 4D Radial MRI Across Organs, Resolutions, and Dynamics Haoyang Pei, Jingjia Chen, Yao Wang, Hersh Chandarana, Li Feng Bernard and Irene Schwartz Center for Biomedical Imaging, Department of Radiology, New York University Grossman School of Medicine, New York, New York, United States of America Impact: The proposed All-in-One DeepGrasp allows for efficient and high-quality, highly-accelerated 4D MRI reconstruction across organs, resolutions, and temporal dynamics, offering significant potential for different clinical applications, including both DCE and non-DCE applications.
16:11		503-03-002. Free-breathing Radial subspace-guided mOtion correction and compressed SENsing (FROSEN) for contrast-enhanced liver imaging Qi Liu, Zihao Chen, Bochao Li, Ran An, Hualing Li, Chenglin Hu, Qiufeng Liu, Yang Yang, Jian Xu United Imaging Healthcare North America, Houston, United States of America Impact: By combining the strengths of prior techniques, FROSEN bridges the image quality gap between free-breathing and breath-hold methods, representing a step toward broader clinical acceptance of free-breathing imaging in abdominal exams.
16:22		503-03-003. AdaSamp: Towards simple, subject-specific Adaptive Sampling for 3D Accelerated MRI Jaehyeok Bae, Zachary Shah, Cagan Alkan, Shreyas Vasanawala, John Pauly, Kawin Setsompop Stanford University, Stanford, United States of America Impact: Our method, AdaSamp, generates simple, subject-specific sampling-mask guided by a fast scout image and tailors k-space coverage to each patient’s spatial-support and anisotropy. It outperforms population-based sampling-masks in reconstruction quality and streamlines practical deployment of subject-adaptive 3D-MRI across diverse anatomies.
16:33		503-03-004. Generative diffusion bridge reconstruction for accelerated motion-compensated free-breathing abdominal MRI Melanie Schellenberg, Richard Do, Ricardo Otazo Memorial Sloan Kettering Cancer Center, New York, United States of America Impact: Diffusion bridge generative AI reconstruction represents a powerful tool for fast motion-compensated abdominal MRI reconstruction, enabling 9-fold acceleration and similar image quality with respect to state-of-the-art motion-resolved imaging.
16:44		503-03-005. PSIRNet: Deep Learning–Based Free-Breathing Rapid-Acquisition Late Enhancement Imaging Arda Atalik, Hui Xue, Daniel Sodickson, Michael Hansen, Peter Kellman Microsoft Research, Redmond, United States of America Impact: PSIRNet reconstructs a phase-sensitive inversion recovery (PSIR) image from a single interleaved IR/PD acquisition thereby significantly shortening the acquisition which is typically 8 to 24 averages. The rapid free-breathing acquisition enables full heart coverage with thinner slices.
16:55		503-03-006. Towards real-time self-gating for fetal cardiac MRI via deep learning-based motion estimation in k-space Aya Ghoul, Christopher Roy, Sarah Nordmeyer, Sergios Gatidis, Thomas Küstner Medical Image and Data Analysis (MIDAS.lab), Department of Diagnostic and Interventional Radiology, Tübingen, Germany Impact: Our deep learning self-gating method enables clinically feasible fetal cardiac MRI without specialized hardware and extensive computation. This may facilitate assessing the fetal cardiac function and development, with potential extensions toward real-time imaging to improve prenatal diagnostics and postnatal care.
17:06		503-03-007. Real4DFlow: Real-time whole-heart 4D flow reconstruction framework from a 5-minute scan using multi-dynamic deep image prior Syed Murtaza Arshad, Salman Pervaiz, Lee Potter, Preethi Subramanian, YINGMIN LIU, Christopher Crabtree, Matthew Tong, Rizwan Ahmad The Ohio State University, Columbus, United States of America Impact: The proposed self-supervised framework enables real-time whole-heart 4D flow imaging, facilitating beat-to-beat variation analysis with significantly improved sharpness, at an unprecedented acceleration. This allows for improved diagnosis of cardiac function, atrioventricular valve disorders, and broader cardiovascular conditions, such as arrhythmias.
17:17		503-03-008. Free-Breathing Double-Beat Exercise CMR with Generative AI for Evaluation of Function, Volumes, and Deformation Manuel Morales, Alexander Schulz, Nicole C.Y. Deng, Kathryn Arcand, Warren Manning, Reza Nezafat Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, United States of America Impact: This study introduces a double-beat, AI-enhanced Ex-CMR sequence to simultaneously capture cardiac volumes and deformation during exercise. By eliminating the need for separate cine and tagging, this approach improves efficiency and enables comprehensive assessment of cardiac performance under physiological stress.
17:28		503-03-009. Simulation of MRI K-Space Motion Corruption for Deep Learning-Based Artifact Detection and Correction Kathryn Lamar-Bruno, Samira Masoudi, Brendan Crabb, Amin Mahmoodi, Alta Steward, Albert Hsiao University of California, San Diego, United States of America Impact: We developed a k-space motion state sampling scheme that yields realistic cardiac and cardiorespiratory artifacts, enabling a multi-task convolutional neural network (CNN) to quantify severity and correct it. This reduces repeat scans thus improving workflow efficiency and enhancing diagnostic reliability.
17:39		503-03-010. Virtual Shimming: A Deep-Learning-Enhanced Dual-Phase bSSFP Imaging Framework for B0-Robust Cardiac Cine MRI Zhuo Chen, Yixin Emu, Haiyang Chen, Zhihao Xue, Sirui Huo, Fan Yang, Sha Hua, Chenxi Hu National Engineering Research Center of Advanced Magnetic Resonance Technologies for Diagnosis and Therapy (NERC-AMRT), Shanghai Jiao Tong University, Shanghai 200240, China Impact: This virtual shimming strategy could allow clinicians to perform subject-agnostic B0-robust bSSFP imaging, even in patients with CIEDs at 3T. It simplifies workflow and potentially improves clinical accessibility of high-quality cardiac MRI.