ISMRM25 - The Perfect Wave: AI-Powered Advances in Image Segmentation

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

The Perfect Wave: AI-Powered Advances in Image Segmentation

Oral

AI & Machine Learning

Monday, 12 May 2025

311

16:00 - 18:00

Moderators: Esin Ozturk-Isik, KyungHyun Sung & Ting Gong

Session Number: O-17

CME Credit

16:00		Introduction Esin Ozturk-Isik
16:12		0254. Exploration of Whole-Body Anatomy in the German National Cohort (NAKO): 3D Segmentation of 55 Structures in 28,969 MRI Scans View Presentation Video L. Fay, Q. Wang, B. Yang, T. Kuestner, S. Gatidis Medical Image and Data Analysis (MIDAS.lab), Department of Diagnostic and Interventional Radiology, University Hospital of Tuebingen, Tuebingen, Germany Impact: This study validates deep learning-based segmentation of the TotalSegmentator model for large-scale MRI analysis (28,969 subjects), showing precise, scalable results. Automated and quality-controlled segmentations demonstrate strong agreement, highlighting its potential to advance research on anatomical structures and health outcomes.
16:24		0255. Automatic segmentation and classification of GBM and SBM using a 3D deep learning model on multiparametric MRI: a multi-center study View Presentation Video M. Wu, J. Luan, H. Wang, X. Wang, X. Liang, C. Zhang, Y. Zhao The Second Hospital of Tianjin Medical University, Tianjin, China Impact: Reading with 3D DL model improves the diagnostic accuracy of radiologists, thereby enhancing patient treatment and prognosis potentially. Considering its promising performance, it is recommended for routine clinical application.
16:36		0256. Multimodal, Multispecies and Pathology Invariant Skullstripping View Presentation Video J. S. Park, J. Ha, S. Tahkur, S. Bakas, E. Garyfallidis Indiana University, Bloomington, Bloomington, United States Impact: This is the first demonstration of a multimodal, multispecies and pathology invariant skullstripping model, only trained on synthetic data with minimal assumptions. Results suggest that with correct assumptions, a single model could be all we need for any skullstripping task.
16:48		0257. On Memory-Based Interactive Deep Learning Models for Cartilage Segmentation in 3D MRIs of the Knee Joint View Presentation Video D. Lopes Ferreira, B. Nunes, X. Zhang, L. Carretero, M. Fung, R. Soni, G. Avinash GE Healthcare, San Ramon, United States Impact: By leveraging memory-based 3D-VFM for morphometric assessment of cartilage through 3DMRIs, we significantly enhance the accuracy and generalizability of the challenging knee soft tissue segmentation, paving the way for more precise measurements of osteoarthritis progression.
17:00		0258. Semi-Supervised Topological Correction View Presentation Video Y. Sun, L. Wang, W. Lin, G. Li, L. Wang University of North Carolina at Chapel Hill, Chapel Hill, United States Impact: The proposed semi-supervised framework addresses topological defects in the segmentation of the brain’s complex folds, providing improved accuracy in cortical analysis and surpassing existing correction methods. This advancement has the potential to enhance studies of neurodevelopmental, neurodegenerative, and psychological disorders.
17:12		0259. Evaluating a Deep Learning Foundation Model for Neuroimaging Segmentation in the Data-Rich and Data-Constrained Settings View Presentation Video K. Nair, Y. Lui, N. Razavian NYU Grossman School of Medicine, New York, United States Impact: While foundation models such as MedSAM have potential for medical segmentation, they currently may not surpass traditional models when using sufficient data. In the data-limited setting, however, they can be useful when extremely little labeled data is available.
17:24		0260. MultiFlowSeg: Unified deep learning model for multi-vessel classification and segmentation of phase-contrast MRI in single ventricle patients View Presentation Video T. Yao, N. St. Clair, G. Miller, A. Zoubian, J. Steeden, R. Rathod, V. Muthurangu University College London, London, United Kingdom Impact: We have developed a pipeline that uses our novel MultiFlowSeg model to automate the extraction, classification, and segmentation of phase-contrast MRI images. It enables rapid, accurate flow quantification for five blood vessels in a single ventricle registry without manual input.
17:36		0261. Pan-Contrast Learning of MRI Segmentation for Healthy and Anomaly Cases: Faithful to Tissue Properties and MR Physics View Presentation Video R. Adams, W. Zhao, S. Hu, W. Lyu, K. Huynh, S. Ahmad, D. Ma, P-T Yap Case Western Reserve University, Cleveland, United States Impact: UBN offers a comprehensive solution for consistent segmentation across all MR image contrasts, vendors, resolutions, sites, preprocessing methods, and age groups.
17:48		0262. Meta-Learning for Robust Medical Image Segmentation: A Gradient-Similarity Reweighting Approach to Mitigate Noisy Labels View Presentation Video A. Al-Fakih, A. Rezk, A. Shazly, K. Ryu, M. A. Al-masni Sejong University, seoul, Korea, Republic of Impact: This advancement addresses noisy data and limited data availability in medical image segmentation, enabling more accurate and reliable predictions.

Similar Session(s)