Cape Town - 2026 ISMRM-ISMRT Annual Meeting and Exhibition - AI for Diagnostic and Prognostic Applications

AI for Diagnostic and Prognostic Applications

Oral

Analysis Methods

Thursday, 14 May 2026

Auditorium 1

16:00 - 17:50

Moderators: Ping Liu & Radka Stoyanova

Session Number: 603-03

No CME/CE Credit

Similar Sessions

This session will cover novel artificial intelligence techniques for diagnostic and prognostic tasks applied to a variety of clinical problems.

16:00		603-03-001. Introduction Radka Stoyanova Jackson Memorial Hospital/University of Miami, United States of America
16:11		603-03-002. SMART-Risk Model to Distinguish Recurrence from pseudoprogression in Brain Tumors: A Large multi-institutional study Dheerendranath Battalapalli, Hyemin Um, Marwa Ismail, Virginia Hill, Sushant Puri, Jennifer S Yu, Lan Lu, Ameya P Nayate, Anthony Higinbotham, Lisa R Rogers, Prateek Prasanna, Mainak Bardhan, Chengnan Li, Mustafa M Basree, Andrew M Baschnagel, Alan McMillan, Ankush Bhatia, Manmeet Singh Ahluwalia, Michael C Veronesi, Pallavi Tiwari University of Wisconsin - Madison, Madison, United States of America Impact: SMART-Risk may enable noninvasive differentiation of tumor recurrence from treatment effects in primary and metastatic brain tumors, reducing biopsies. By integrating spatial, textural, and morphological descriptors beyond visual MRI assessment, SMART-Risk captures subtle post-treatment lesion attributes to inform clinical decision-making.
16:22		603-03-003. Biology-Informed Nomogram for Risk Stratification of Glioblastoma Survival Using MRI-based Hemodynamic Heterogeneity Features Junfeng Zhang, Xiaoxiao Ma, xiaojun yu, Yujue Zhong, Jian Hu, Xin Lou The First Medical Center, Chinese PLA General Hospital, Beijing, China Impact: Refined risk stratification to reduce survival imbalances between study arms in glioblastoma trials is critical. We developed an online-accessible interactive nomogram incorporating imaging-clinical variables that enables risk stratification with favorable biological interpretability for guiding individualized patient management and trial design.
16:33		603-03-004. A Hierarchical Multi-task Learning Framework for Brain Tumor Classification Using Multi-modal MRI Qianqian Zheng, Yueqi Yang, Shuang Li, Yuchi Tian, Xiaorui Su, Xiaoyun Liang, Shuang Tang, Guangliang Ju, Qiang Yue West China Hospital of Sichuan University, Chendu, China Impact: This work proposed a hierarchical diagnostic model achieving high accuracy in multi-level brain tumor classification. Its segmentation-guided classification improves fine-grained prediction and supports clinically relevant, interpretable outcomes.
16:44		603-03-005. Behavior Score Prediction in Resting-State Functional MRI for Alzheimer’s Disease Spectrum by Deep State Space Modeling Javier Salazar Cavazos, Maximillian Egan, Benjamin Hampstead, Scott Peltier University of Michigan, Ann Arbor, United States of America Impact: The NeuroMamba model improves behavior score prediction for rs-fMRI enabling insights of key brain regions tied to MoCA, memory, and language metrics that can pave the way for early intervention techniques and monitoring. Future research can explore pathology-related task-based fMRI.
16:55		603-03-006. Interpretable MRI Automated Machine Learning Model for Predicting Response to Combination of Antiangiogenic and Immunotherapy Huihui Wang, Ying Xu, Sicong Wang, Lizhi Xie, Feng Ye, Xinming Zhao National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China Impact: An interpretable MRI-based automated machine-learning model (HAIRS) accurately predicts response to anti-angiogenic plus immunotherapy in HCC across multicenter training, external, and prospective cohorts, providing a transparent decision-support tool that may guide patient selection and trial enrichment strategies.
17:06		603-03-007. Context is everything: Reducing false positives in longitudinal health assessment using deep learning with prior information Lavanya Umapathy, Patricia Johnson, Tarun Dutt, Angela Tong, Madhur Nayan, Hersh Chandarana, Daniel Sodickson Bernard and Irene Schwartz Center for Biomedical Imaging, Department of Radiology, New York University Grossman School of Medicine, New York, New York, United States of America Impact: A reduction in false positive rates using individualized prior context without degrading sensitivity could offer a pathway to expand longitudinal health-monitoring programs to large populations with comparatively low baseline risk of disease, leading to earlier detection and improved health outcomes.
17:17		603-03-008. Fusing Medical History Trajectories and Multi-modal Image Features for Disease Risk Prediction Zian Wang, Haoyang Zhang, Lizhen Lan, Yan Li, Yajing Zhang, Chengyan Wang School of Computer Science, Fudan University, Shanghai, China Impact: This study moves beyond single-timepoint phenotyping by modeling dynamic disease trajectories and integrating them with imaging phenotypes, enabling more holistic patient assessment for improved disease risk prediction. Such fusion enables an average increase of 13.7% risk prediction performance against references.
17:28		603-03-009. Body fat distribution predicts cardiometabolic risk in healthy non-obese individuals: an opportunistic screening approach Balazs Bogner, Matthias Jung, Marco Reisert, Juliane Maushagen, Susanne Rospleszcz, Fabian Bamberg, Jana Taron, Jakob Weiß University Medical Center Freiburg — Department of Diagnostic and Interventional Radiology, Freiburg, Germany Impact: Automated MRI-derived VAT/SAT ratio reveals hidden cardiometabolic risk in apparently healthy individuals missed by conventional measures, providing rationale for opportunistic screening from routine clinical imaging to identify individuals who could potentially benefit from earlier health interventions.
17:39		603-03-010. Machine Learning Based on Multiparametric MRI Radiomic for Pathology Aggressive Prediction in Clear Cell Renal Cell Carcinoma Jie Zhan, Lei Sun, Enming Cui, Zhitao Yang, Jiankun Dai, Xin Zhen, Ruimeng Yang Guangzhou First People’s Hospital, Guangzhou，Guangdong, China Impact: For the first time, our study demonstrated that T2WI-derived radiomic features are superior for non-invasively predicting aggressive pathology in ccRCC. This suggests that T2WI should be prioritized in radiomics pipelines for ccRCC risk stratification, potentially guiding personalized treatment decisions.