ISMRM25 - Carving AI Currents in Image Synthesis

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

Carving AI Currents in Image Synthesis

Oral

AI & Machine Learning

Thursday, 15 May 2025

310 (Lili-u Theater)

08:15 - 10:15

Moderators: Zhaolin Chen & Xiaoping Wu

Session Number: O-11

No CME/CE Credit

08:15		1114. Predicting Delayed Phase Contrast-Enhanced MR Images from Early Phase Contrast-Enhanced MR Images Using Deep Learning-Based Iterative Network View Presentation Video W. Chung, J. Kang, G. E. Park, S. H. Kim, Y. Nam Hankuk University of Foreign Studies, Yongin-si, Korea, Republic of Impact: By enabling dynamic contrast prediction in breast MRI, our method aids in the characterization of enhancement patterns in breast tissue using only early phase post-contrast images. This approach potentially reduces scan times for dynamic contrast-enhanced MR applications.
08:27		1115. A Deep Learning Approach to Longitudinal Infant MRI Synthesis View Presentation Video Y. Fang, H. Xiong, J. Huang, F. Liu, X. Cai, Z. Shen, H. Zhang, Q. Wang ShanghaiTech University, Shanghai, China Impact: This framework enables accurate tracking of infant brain development by filling missing MRI data, aiding in the creation of developmental atlases, and supporting early detection of disorders. It may thus advance both neurodevelopmental research and clinical interventions.
08:39		1116. Towards Metadata-customized Brain MR image Synthesis for Disease Diagnosis View Presentation Video Y. Wang, H. Xiong, K. Sun, S. Bai, Z. Ding, Q. Wang, Q. Liu, D. Shen Shanghaitech University, Shanghai, China Impact: Our general multimodal MRI synthesis foundation model is capable of quickly and cost-effectively providing metadata-tailored multiple MR sequences, enabling clinicians and researchers to customize the desired MR images using this convenient AI technology, thereby enhancing diagnostic precision and efficiency.
08:51		1117. Controllable Magnetic Resonance Image Contrast Adjustment via Sequence Parameter-Driven Network View Presentation Video H. Jang, H. Kim, Y. Song, D. Hwang Yonsei University, Seoul, Korea, Republic of Impact: This method adjusts the contrast of MR images based on MR sequence parameters without requiring additional scans. This approach has the potential to reduce the scan time needed to acquire multi-contrast MR images.
09:03		1118. Synthesizing brain-originated realistic diffusion-weighted MRI signal for in silico experiments View Presentation Video T. Pieciak, S. Aja-Fernández, A. Tristán-Vega Universidad de Valladolid, Valladolid, Spain Impact: The proposal models brain-originated characteristics and enables diffusion-weighted data synthesis reflecting the observed MRI signal. Compared to previous solutions, which fix brain characteristics or draw them randomly, our approach realistically varies signal properties based on what is observed.
09:15		1119. Multi-Contrast MR Image Synthesis with Episodic State-Space Modeling View Presentation Video Ö. Atlı, B. Kabas, F. Arslan, A. Demirtas, M. Yurt, O. Dalmaz, T. Cukur Bilkent University, Ankara, Turkey Impact: The extended scope of multi-contrast protocols enabled through I2I-Mamba may facilitate comprehensive MRI exams in numerous applications, including assessment of pediatric and elderly individuals in need of rapid scans given limited motor control and vulnerability to toxicity from contrast agents.
09:27		1120. Synthesizing Full-dose FDG Brain PET from MRI With and Without Ultralow-dose PET using Deep Learning Diffusion Models in Patients with Epilepsy View Presentation Video J. Wu, J. Ouyang, M. Khalighi, G. Zaharchuk Stanford University, Stanford, United States Impact: This study suggests the possibility to use generative AI approaches to massively reduce dose levels for FDG PET brain studies. Further work will leverage 3D patch-based approaches can improve the performance and slice consistencies.
09:39		1121. Deep Learning-Based High b-Value Image Synthesis: Application to SANDI Microstructure Map Prediction View Presentation Video R. Zheng, Y. Li, H. Zhang, B. Zhang, X. Xia, Z. Tang, C. Wang, Y. Chu, H. Zhang, C. Wang, H. Li, H. Wang Fudan University, Shanghai, China Impact: This study explores the feasibility of replacing real high b-value images with synthesized images generated by deep learning models, which holds promise for transfer to other MRI systems with lower gradient performance, thereby expanding the application scope of SANDI.
09:51		1122. Physics-informed Latent Diffusion Model multi-echo chemical shift-encoded liver MRI generation View Presentation Video J. Meneses, Y. George, C. Hagemeyer, Z. Chen, S. Uribe Pontificia Universidad Católica de Chile, Santiago, Chile Impact: We successfully generated realistic multi-echo liver MR images and relevant quantitative maps to train deep learning models for PDFF estimation. The combination of limited real samples and numerous synthetic images for training enabled an improved performance compared to real-only datasets.
10:03		1123. Rescuing Incomplete MR Data: Anatomy Imputation of Restricted Field of View Images Using Multi-Contrast MR Images View Presentation Video S. Hays, S. Remedios, L. Zuo, J. Zhang, A. Carass, E. Mowry, S. Newsome, J. Prince, B. Dewey Johns Hopkins University, Baltimore, United States Impact: Our results impact researchers handling diverse, inconsistent imaging datasets with variable field-of-view acquisitions. This approach enables the analysis of previously unusable data by imputing missing regions using multi-contrast information, making them suitable for meaningful clinical or research outcomes.

Similar Session(s)