Cape Town - 2026 ISMRM-ISMRT Annual Meeting and Exhibition

AI Methods

Traditional Poster

Acquisition & Reconstruction

Wednesday, 13 May 2026

Traditional Posters | Exhibition Hall

09:15 - 10:10

Session Number: 570-03

No CME/CE Credit

Similar Sessions

AI methods used in various ways to improve MRI are presented in this session.

		570-03-180. Predicting T1rho Map with PD- and T2-weighted MRI for Knee Joint Junru Zhong, Chaoxing Huang, Ziqiang Yu, Lu Wen, Hongjian Kang, Qianxue Shan, Fuqian Guo, Queenie Chan, James Griffith, Weitian Chen The Chinese University of Hong Kong, Hong Kong, China Impact: We demonstrated the capability of deep learning neural networks to predict T1rho maps without requiring actual T1rho scans, which could significantly reduce scan time and enable retrospective analysis of valuable historical clinical data.
		570-03-181. Zero‑Shot Low‑Field MRI Quality Enhancement Using a Noise Level Adaptive Diffusion Model (Nila) Jiacai Cai, Nanxiong Liu, Shoujin Huang, Yuwan Wang, Yansong Bu, Zihao Wang, Shaojun Liu, Min Wang, Mengye Lyu Shenzhen Technology University, Shenzhen, China Impact: This method boosts readability of routine Low‑field scans using only high‑field training data, requires no paired data or retraining.
		570-03-182. Deep Learning Super-Resolution for T1-Weighted Magnetic Resonance Imaging at 0.5T Ying Yang, Diego Martinez, Amgad Louka, Alexander Mertens, Ian Connell University of Toronto, Toronto, Canada Impact: The use of super-resolution techniques coupled with 0.5T MRI can be used to reduce scan time and boost SNR, showing potential in patient applications that may not be achievable at higher field strengths.
		570-03-183. Deep Learning-Augmented SENSE/GRAPPA Parallel Imaging for Enhanced Image Quality and Diagnostic Accuracy in Femoroacetabular qi tong Liu, Chen Zhang Beijing Chaoyang Hospital, Capital Medical University, beijing, China Impact: To optimize Femoroacetabular Impingement (FAI) MRI protocols, explore the application value of integrating Deep Learning (DL) with multimodal parallel imaging technologies to facilitate clinical popularization, and open up new directions for AI-based musculoskeletal imaging.
		570-03-184. Liver Cirrhosis Visual Severity Estimation From MRI With Deep Learning Jun Zeng, Halil Ertugrul Aktas, Elif Keles, Alpay Medetalibeyoglu, Matthew Antalek, Amir Borhani, Daniela Ladner, Gorkem Durak, Ulas Bagci Chongqing University of Posts and Telecommunications, Chongqing, China Impact: This study introduces a large-scale, multi-sequence MRI deep learning framework for automated visual estimation of liver cirrhosis severity. Accurately stratifying patients early using MRI images will lead to better patient management and ultimately better outcomes.
		570-03-185. 2D to 3D MR Image Super-Resolution using Cross-Contrast Guidance Zheng Zhang, Zechen Zhou, Lei Xiang, Ajit Shankaranarayanan, Xinyu Song, Yuehua Li Subtle Medical Inc, Menlo Park, United States of America Impact: This work introduces a cross-contrast 3D super-resolution framework that reconstructs diagnostic-quality high-resolution MRI from rapid 2D scans, enabling significant scan-time reduction while preserving fine anatomical detail, improving lesion visibility, and enhancing the clinical feasibility of accelerated multi-contrast brain imaging.
		570-03-186. Deep learning-accelerated fat suppressed T2-weighted imaging of the breast: Faster acquisition with comparable image quality Chenxi Zhang, Yuanfeng Wei, Hao dong Qin, Marcel Dominik Nickel, Mingjue Jian, Li Ding, Xiaomei Li, Yingrui Huangyingrui, Chenggong Yan Nanfang Hospital, Southern Medical University, Guangzhou 510515, China Impact: Deep learning-accelerated T2-weighted imaging of the breast significantly reduces scan time while maintaining diagnostic quality, offering a feasible solution to optimize breast MRI workflow.
		570-03-187. Neural Networks for Dictionary-Free MR Fingerprint Matching in the Brain: A Systematic Review Ela Kanani, Geoff Parker, Elizabeth Powell University College London, London, United Kingdom Impact: ML for MRF parameter is a promising alternative to dictionary matching. However, standardised evaluation protocols and validation in disease populations are essential to determine which architectures best suit different sequences and clinical applications.
		570-03-188. UniCMR: Enhancing Reconstruction Generalization for Cross-Center Cardiac MR via A Cascaded Network and DFU Module Yipin Deng, Yuyang Li, Zijian Zhou, Yining Wang, Yuan Feng, Peng Hu School of Biomedical Engineering, ShanghaiTech University, Shanghai, China Impact: The proposed cascaded network, incorporating DFU modules to adapt to multi-center inputs, offers a robust solution to the domain shift issue in MRI reconstruction and expands its cross-center generalization.