ISMRM25 - AI: Methods & Applications

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

AI: Methods & Applications

Power Pitch

AI & Machine Learning

Thursday, 15 May 2025

Power Pitch Theatre 1

13:15 - 15:15

Moderators: Ashish Saxena & Esin Ozturk-Isik

Session Number: PP-05

No CME/CE Credit

13:15		Screen Number: 1 1283. Microstructure Quantification by Q-Space Trajectory Imaging via Unrolled Neural Networks: Exploring Model Generalizability View Presentation Video J. Yu, O. Gödicke, F. Laun, M. Ladd, T. Kuder German Cancer Research Center (DKFZ), Heidelberg, Germany Impact: The developed network accelerates model training with higher fidelity than state-of-the-art machine learning methods. The RP approach enhances model robustness and supports generalizability, facilitating on-the-fly QTI microstructural estimation across different acquisition protocols, which may improve clinical utility.
13:17		Screen Number: 2 1284. Beyond segmentation: an uncertainty-aware end-to-end approach to functional lung image quantification View Presentation Video J. Astley, H. Marshall, L. Smith, A. Biancardi, G. Collier, J. Wild, B. Tahir University of Sheffield, Sheffield, United Kingdom Impact: Direct prediction of ¹²⁹Xe-MRI lung ventilation metrics using a multi-modality dual-channel convolutional neural network for end-to-end functional lung image quantification. Uncertainty quantification is utilized to improve trust and reliability in predictions and facilitate streamlined triaging in functional lung imaging workflows.
13:19		Screen Number: 3 1285. Biological mechanisms underlying prognostic radiomic models based on dynamic susceptibility contrast perfusion-weighted imaging in gliomas View Presentation Video J. Yan, C. Zhang The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China Impact: Our work presents an accurate and biologically meaningful tool for survival prediction in gliomas, facilitating personalized clinical decision-making through a non-invasive approach.
13:21		Screen Number: 4 1286. Diffusion-Weighted MRI Radiomics Model for Differentiating High-Grade Gliomas in Non-Enhancing (Low-Grade Appearance) Gliomas View Presentation Video Y. Liu, Y. Liang, Z. Chen, Y. Wang Chinese PLA General Hospital, Beijing, China Impact: The radiomics model based on diffusion MRI can non-invasively differentiate between high-grade and low-grade gliomas in non-CE adult diffuse gliomas,providing valuable support for decision-making in clinical surgical strategies and prognostic management.
13:23		Screen Number: 5 1287. Explainable automated image quality assessment for magnetic resonance imaging through prediction of defect maps View Presentation Video V. Saksena, S. Arroyo-Camejo, R. Schneider Siemens Healthineers, Erlangen, Germany Impact: Clinical adoption of automated image quality assessment methods can help technologists in automatically monitoring image quality. Meaningful visual explanations offered by the proposed approach could help in building trust in the method and allow fast false positive detection by technologists.
13:25		Screen Number: 6 1288. Image quality evaluation of multi-sensitivity diffusion-weighted imaging for rectal cancer with small field of view based on deep learning X. Zhang, Q. Xu, L. Guo, Y. Shi, A. Dong, D. Pylypenko The First Huai 'an Hospital Affiliated to Nanjing Medical University, Huai 'an, China, Huai 'an, China Impact: This study shows that DLR-based FOCUS-MUSE DWI enhances image quality, rectal contour, lesion visibility, SNR, and CNR, suggesting its potential to improve colorectal cancer staging accuracy.
13:27		Screen Number: 7 1289. Biological age assessment in 100,000 whole-body MRI of the German National Cohort (NAKO) and UK Biobank View Presentation Video V. Ecker, B. Yang, S. Gatidis, T. Küstner University Hospital of Tübingen, Tübingen, Germany Impact: Our imaging-based multi-organ prediction of biological age from whole-body MRI of 100,000 participants in the UK Biobank and NAKO cohorts provides an important foundation to investigate aging patterns and influencing factors.
13:29		Screen Number: 8 1290. Exploring the possibilities with deep learning to compute shape measures of the brain's white matter connections View Presentation Video Y. Lo, Y. Chen, D. Liu, J. H. Legarreta, L. Zekelman, J. Rushmore, F. Zhang, Y. Rathi, N. Makris, A. Golby, W. Cai, L. O'Donnell Harvard Medical School, Boston, United States Impact: We investigate the possibility of deep learning to compute shape measures of the brain's white matter connections without intermediate steps to convert geometric tractography streamline data to an image data representation using a voxel grid with our novel framework, TractShapeNet.
13:31		Screen Number: 9 1291. Applying Integrating Event-Based Models and Deep Learning to Predict Baseline Diagnosis and Tau Progression in Alzheimer's Disease D. Ma, R. Sandell, A. Raj UCSF, San Francisco, United States Impact: Our method enhances understanding of inter-subject heterogeneity in AD, bridges the information misalignment in Tau-PET and MRI, and supports precision treatment by predicting individual tau progression and seeding patterns with only baseline MRI and demographic inputs.
13:33		Screen Number: 10 1292. A Self-Supervised Deep Learning Method for Accelerated Quantification of High-Resolution Short-TE spectroscopic MRI Datasets View Presentation Video A. Giuffrida, S. Sheriff, B. Weinberg, L. Cooper, B. Soher, M. Treadway, A. Maudsley, H. Shim Emory University, Atlanta, United States Impact: A deep learning method for accelerated quantification of spectroscopic MRI datasets with short echo time, NNFit, achieved competitive quantitative performance in comparison to a standard parametric modelling spectral analysis method with greater computational efficiency.
13:35		Screen Number: 11 1293. A Generalized Cascaded Self-Supervised Registration Pipeline with Physics-Informed Learning for Enhanced Quantitative Cardiac MRI View Presentation Video X. Li, Y. Zhang, L-T Huang, H. Chang, T. Niendorf, K-L Nguyen, M-C Ku, Q. Tao, H-J Yang Cedars-Sinai Medical Center, Los Angeles, United States Impact: The cascaded self-supervised pipeline with a physics-informed module offers a scalable framework to facilitate accurate and efficient image registration for image contrast modulation following multiple physical and physiological models.
13:37		Screen Number: 12 1294. Quantifying Cerebral Small Vessel Geometry with Self-Supervised Shape-Aware 3D U-Net on 7T TOF MRI: Associations with Age and Cognitive Function View Presentation Video J. Tang, Z. Deng, E. Joe, H. Chui, Y. Shi, L. Yan Northwestern University, Chicago, United States Impact: Quantitative morphological measures of cerebral small vessels can be effectively segmented using deep learning, which can be used as sensitive markers of aging, cognitive dysfunction, and brain vascular function
13:39		Screen Number: 13 1295. Towards confounding-free and causal predictions: Mitigation of Multiple Spurious Correlations in Deep Learning-Based MRI analysis View Presentation Video L. Fay, H. Reguigui, B. Yang, S. Gatidis, T. Kuestner Medical Image and Data Analysis (MIDAS.lab), Department of Diagnostic and Interventional Radiology, University Hospital of Tuebingen, Tuebingen, Germany Impact: Our novel framework, MIMM-X, an extension of our previous model (MIMM), is able to remove multiple spurious correlations in MRI, ensuring causal predictions based on task-relevant features. This approach improves generalization for data across various MR scanners and patient demographics.
13:41		Screen Number: 14 1296. WITHDRAWN
13:43		Screen Number: 15 1297. Towards Reliable Deep Learning: Feature-Based Out-Of-Distribution Detection for Brain Morphometry View Presentation Video T. Di Noto, L. Bacha, K. Prabhu M, V. Dunet, A. Jantarato, M. Vaneckova, E. Le Bars, N. Menjot de Champfleur, P. Venkategowda, B. Maréchal Siemens Healthineers International AG, Lausanne, Switzerland Impact: We experiment feature-based Out-Of-Distribution (OOD) detection to identify problematic scans for which segmentation results might be unreliable. While Near-OOD remains an area of future improvement, our approach is effective for the majority of use cases and adds negligible computation time.
13:45		Screen Number: 16 1298. APGC Net: Unsupervised Cross-Modality Adaptation for Multi-organ Segmentation in TAO via adaptive pseudo-label-guided contrastive learning View Presentation Video Y. Sun, X. Zhou, M. Deng, C. Chen, Q. Dou, K. Chan, K. Chong, W. Chu The Chinese University of Hong Kong, Hong Kong, Hong Kong Impact: The impact lies in the technical innovation to ensure consistently accurate segmentation of each organ involved in TAO on multi-modal MRI, which is beneficial to alleviating the burden of manual labeling and reducing observer variability.
13:47		Screen Number: 17 1299. An Imageless Magnetic Resonance Diagnosis procedure for fast and affordable screening and follow-up View Presentation Video P. García Cristóbal, A. González Cebrián, F. Galve, V. Van Der Valk, E. Ilıcak, M. Staring, A. Webb, J. Alonso Instituto de Instrumentación para Imagen Molecular, CSIC, Universitat Politècnica de València, Valencia, Spain Impact: The use of Imageless MR sequences, combined with deep-learning methods, could offer a rapid, cost-effective screening technique suitable for large population-wise deployment. In simulations we show how white matter lesions could potentially be detected and characterized.
13:49		Screen Number: 18 1300. Deep-PRL: a deep learning network for the identification of paramagnetic rim lesions in multiple sclerosis View Presentation Video F. Spagnolo, A. Bhardwaj, P. Gordaliza, P-J Lu, M. Ocampo-Pineda, M. Bach Cuadra, X. Chen, B. Ayci, A. Cagol, V. Andrearczyk, A. Depeursinge, C. Granziera University Hospital Basel and University of Basel, Basel, Switzerland Impact: These results represent a significant step towards the integration of an AI tool to assist clinicians in the identification of PRLs, thereby improving the clinical management of pwMS.
13:51		Screen Number: 19 1301. Prediction of treatment response in a longitudinal glioblastoma dataset using deep learning View Presentation Video A. Matoso, C. Passarinho, M. Loureiro, J. Moreira, P. Figueiredo, R. Nunes Instituto Superior Técnico, Universidade de Lisboa, Lisbon, Portugal Impact: This work sheds light on the deep learning strategies for predicting treatment response in glioblastomas, highlighting the approaches that perform best, thus providing valuable insights into the optimization of the prognostic accuracy of these models.
13:53		Screen Number: 20 1302. Deep learning-based fetal brain extraction method for in utero diffusion MRI View Presentation Video Z. Zhang, J. Li, Y. Chen, Y. Feng, X. Zhang Southern Medical University, Guangzhou, China Impact: The proposed method can significantly streamline the tedious annotation process and improve segmentation accuracy, contributing to a fast and accurate post-processing pipeline.
13:55		Screen Number: 21 1303. Low latency in-line segmentation of cardiac structures for real-time cardiac MRI View Presentation Video K. Lee, Y. Tian, K. Nayak University of Southern California, Los Angeles, United States Impact: We demonstrate in-line low-latency segmentation with comparable LV and MYO Dice score to state-of-the-art-methods using a semantic segmentation model.
13:57		Screen Number: 22 1304. Decoding heterogeneity in solitary HCC: OATP-dependent HBP-MRI radiomics predicts microvascular invasion & prognosis. Y. can, Z. Yang The Affiliated Cancer Hospital of Harbin Medical University, harbin, China Impact: This study integrates multiple machine learning methods to predict MVI in HCC, revealing an ECM-OATP1B3 correlation and enhancing the biological interpretability of the radiomics model for prognostic assessment.
13:59		Screen Number: 23 1305. Development and Validation of an MRI-Based Predictive Model for Preoperative Extrathyroidal Extension in Papillary Thyroid Carcinoma View Presentation Video B. Chen, Y. Song, H. Wang, L. Tang, X. Xie, A. Mao, P-Y Wu, Q. Chen, B. Song Department of Radiology, Minhang Hospital, Fudan University, Shanghai, China Impact: Our nomogram model, incorporating age, protrude_value, and ADC_Best_rate, effectively predicted preoperative ETE in PTC patients, thereby aiding surgeons in optimizing therapeutic decision-making. ADC_Best_rate demonstrated potential as a robust indicator in MRI functional imaging.
14:01		Screen Number: 24 1306. Prediction of High and Low Expression of Tumor-Infiltrating Lymphocytes in Breast Cancer Using MRI Features Combined with Molecular Subtypes View Presentation Video J. Zhou, Y. Zhang, Y-L Liu, J-H Chen, M. Wang, M-Y Su The First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China Impact: MRI features could predict high vs. low TILs expression. The three molecular subtypes (HR+/HER2-, HER2+, triple-negative) had distinctly different TILs, and more sophisticated models by combining MRI features with clinical and histological information could improve the TILs prediction accuracy.
14:03		Screen Number: 25 1307. Beyond Brain Age: QS-GAP as a Quantile-Based Tool for Detecting Neurodegenerative Aging Patterns View Presentation Video R. Navarro-Gonzalez, R. de Luis-Garcia, S. Aja-Fernández Universidad de Valladolid, Valladolid, Spain Impact: QS-GAP offers a refined, personalized approach to assessing brain aging, enhancing classification between AD and CN individuals. By aligning aging patterns with cohort norms, it supports more accurate diagnostics and advances neurodegenerative research, particularly in individualized brain health assessments.

Similar Session(s)