ISMRM25 - AI-Based Image Recon, Enhancement & Analysis

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

AI-Based Image Recon, Enhancement & Analysis

Traditional Poster

Tuesday, 13 May 2025

Building: Room: Exhibition Hall

16:45 - 17:45

Session Number: T-04

No CME/CE Credit

		5139. Deep learning-based residual-subtractive denoising with latent mapping of noise: Application to Z-spectra PDF Poster S. Mohammed Ali, N. Yadav, R. Wirestam, P. van Zijl, J. Prasuhn, L. Knutsson Lund University, Lund, Sweden Impact: Denoising is crucial for usability of Z-spectra from CEST-images. Our developed DL-based solution is designed to map noise in latent space and subsequently remove it in a residual fashion allowing for an increased potential to recover overshadowed signal in Z-spectra.
		5140. Frequency and Phase Correction of MRS Data by Leveraging Sequential Patterns along the Transient Dimension C. Wu, K. Igwe, J. Guo Columbia University, New York, United States Impact: The 2D-FPC method proposed follows an unsupervised learning approach that allows it to be directly applied in vivo dataset without the need for pre-training. The simultaneous processing of all transients allows the algorithm to capture long-range patterns in the data.
		5141. A Physics-Informed Convolutional Neural Network to Estimate Intravoxel Incoherent Motion Parameters in the Liver PDF Poster M. Brown, J. Vasquez, A. Moody, M. Abdul-Ghani, R. DeFronzo, J. Blangero, G. Clarke University of Texas Health Science Center San Antonio, San Antonio, United States Impact: This study showed that CNNs improve the repeatability of D* and D estimates in the liver, though it remains unclear if the within-subject variability of IVIM parameters is sufficient to accurately differentiate fibrosis stages.
		5142. Accelerating 7T Quantitative MRI with Self-Supervised Few-Shot Deep Learning PDF Poster R. Qiu, M. Safari, Z. Eidex, S. Wang, M. Hu, H. Mao, E. Middlebrooks, X. Yang Emory University, Atlanta, United States Impact: The proposed DL approach for accelerating 7T qMRI reduces scan times without compromising image quality, facilitating broader adoption of high-field MRI. Its generalizability with limited training data enhances advanced neuroimaging accessibility and efficiency, contributing to better clinical utility.
		5143. Evaluation of AI-Assisted Compressed Sensing for Accelerated Shoulder FACT MRI Sequences M. Yang, C. Tian, M. Shao, Z. Wang, Y. Guo, Z. Lin, X. Yang, X. Zhang Department of Radiology, Nanjing Drum Tower Hospital, Affiliated Hospital of Medical School, Nanjing University, No. 321, Zhongshan Road, Nanjing, Jiangsu, 210008, China , Nanjing, China Impact: The study demonstrates that ACS improves the clinical efficiency of FACT sequences in shoulder imaging by reducing scan times without affecting quantitative metrics or image quality.
		5144. DeepRelaxo: A Generalizable Self-supervised Method for Brain R2* Mapping PDF Poster S. Prima, Z. Xiong, A. H. Wilman, H. Sun The University of Queensland, Brisbane, Australia Impact: Self-supervised training removes the need for real-world scans and allows DeepRelaxo adaptable across diverse clinical sites. Improved shorter echo reconstructions facilitates significantly scan time reduction.
		5145. Deep learning-based sub-pixel level motion estimation and correction in the context of the bSSFP-based radial fMRI PDF Poster F. Makhsousi, S. Ghaffarzadeh, B. Feizifar, A. Nasiraei Moghaddam Institute for Research in Fundamental Sciences, Tehran, Iran (Islamic Republic of) Impact: In this study, the translational and rotational motion at the sub-pixel level was estimated and corrected using the kspace. It may result in more accurate detection of neural activity and potentially improve fMRI research.
		5146. QSM-RimDS: A highly sensitive paramagnetic rim lesion detection and segmentation tool for multiple sclerosis lesions H. M. Luu, I. Kovanlikaya, T. Vu, M. Sisman, P. Spincemaille, Y. Wang, S. Gauthier, T. Nguyen Weill Cornell Medicine, New York, United States Impact: QSM-RimDS has the potential to replace human readers for the time-consuming PRL detection and segmentation task while improving reproducibility.
		5147. Line-based Instance Segmentation for Robust and Automatic MRI Spine Scan Planning PDF Poster M. He, X. Hao, L. Guo, M. Song, L. Chen, B. Qiu University of Science and Technology of China, Hefei, Anhui, China Impact: Our method improved the accuracy and speed of spine positioning, providing more precise spinal transection scanning results. This can help doctors in diagnosing lumbar spine diseases more effectively and improve the diagnosis experience of patients.
		5148. Evaluation of the Automatic Segmentation of Aorta considering Aortic Diameter Measurements from Native MRI Angiographic Images PDF Poster Y. Zhou, A. Barrera-Naranjo, T. Decourselle, D. Marin-Castrillon, B. Presles, M. Delcey, O. Bouchot, J-J Christophe, A. Lalande ICMUB laboratory, CNRS 6302, University of Burgundy, Dijon, France Impact: We validated an automatic method for the aortic segmentation from native angiography using the diameter measurements, considering the data from medical reports as the ground truth, instead of traditional metrics like the Dice coefficient or Hausdorff Distance.
		5149. Deep Learning based Tissue Specific MRI-IVIM Parameter Estimation PDF Poster J. Liu, Z. Yang, R. Moreno KTH Royal Institute of Technology, Stockholm, Sweden Impact: IVIM model has gained momentum recently, especially in the field oncology. Our study improve the parameter estimation performance for IVIM model, which is important for tumor and tumor type prediction.
		5150. K-space Interpolation using Deep Koopman Autoencoders. PDF Poster W. Ben Salah, S. McElroy, J. Shapey, S. Ourselin, C. Bergeles, R. Neji King's College London, London, United Kingdom Impact: This work introduces an interpretable neural network for k-space interpolation, enabling good reconstruction quality and offering avenues for extensions to enable autoencoder-based scan-specific denoising.
		5151. DMnet: A Reliable Magnetic Resonance Spectroscopy Quantification Network and the Verification on 5T scanners PDF Poster Z. Tu, J. Zhang, Y. Yang, D. Guo, X. Qu Xiamen University, Xiamen, China Impact: The DMnet can quantitate MRS more quickly and robustly, even in scenarios with lower SNRs. This method has been integrated into the CloudBrain-MRS platform for convenient one-click access by healthcare professionals, further aiding clinical treatments.
		5152. The Impact of Training Data on MRS Metabolite Quantification with Deep Learning PDF Poster Z. Ma, O. Karakus, S. Shermer, F. Langbein Cardiff University, Cardiff, United Kingdom Impact: This study underscores the importance of training data quality in deep learning for MRS. By demonstrating the impact of noise model realism, it provides insights for developing more accurate metabolite quantification models, potentially improving clinical diagnosis and monitoring neurological disorders.
		5153. Convex implicit multi-scale (i-MuSE) energy framework: bridging compressed sensing and diffusion models PDF Poster J. Rikhab Chand, M. Jacob University of Virginia, Charlottesville, United States Impact: i-MuSE offers a memory-efficient alternative to unrolled models while guaranteeing convergence, offering better generalization performance, and facilitating fast optimization algorithms. It can also perform posterior sampling, like diffusion models, to estimate uncertainty.
		5154. Deep learning enabled motion detection in quantitative macromolecule proton faction mapping in the liver PDF Poster Q. Shen, V. Wong, J. Zhong, H. Kang, Z. Yu, Q. Chan, W. Chu, W. Chen CU Lab of AI in Radiology (CLAIR), Department of Imaging and Interventional Radiology, The Chinese University of Hong Kong, Hong Kong, Hong Kong Impact: Our approach enables automated motion detection of the liver during MPF-SL scan. It can improve reliability of parameter quantification by either discarding unreliable measurements retrospectively or prompt data recollection prospectively during scanning.
		5155. MIRAGE: MR Image Replication without Repeated Data Acquisition Using Generative Model PDF Poster Y-J Jeong, C-H Oh Korea University, Seoul, Korea, Republic of Impact: MIRAGE offers a promising solution for efficient SNR enhancement, with potential applications across various configurations, including different anatomical regions, imaging protocols, field strengths, and x-nuclei imaging.
		5156. Deep Learning-based Fast Calculation of Diffusion Tensor Distribution Parameters J. Zhou, Z. Zhu, F. Zong, X. Deng, P. Or, D. Topgaard School of Artificial Intelligence, Beijing University of Post and Telecommunication, Beijing, China Impact: This study introduces deep learning for DTD parameter computation, overcoming computational complexity and spatial continuity issues of Monte Carlo methods, with potential for clinical translation across various diseases.
		5157. Transfer Learning for Segmentation of the Whole Prostate and Intraprostatic Lesions in Multi-Parametric MRI A. Ali, L. Muralidharan, S. Punwani, A. Retter, K. Shmueli University College London, London, United Kingdom Impact: A nnU-Net network trained on large public datasets, then fine-tuned with a small clinical dataset improved whole-prostate segmentation. This network will facilitate processing requiring whole-prostate masks, such as Quantitative Susceptibility Mapping, and could potentially reduce radiological workload or automate quantification.

Similar Session(s)