Cape Town - 2026 ISMRM-ISMRT Annual Meeting and Exhibition - Image Classification for Brain Cancer

Image Classification for Brain Cancer

Digital Poster

Analysis Methods

Wednesday, 13 May 2026

Digital Posters Row D

13:40 - 14:35

Session Number: 563-03

No CME/CE Credit

Similar Sessions

This session will highlight recent advances in brain cancer imaging and classification, with a focus on quantitative and AI-based approaches. Contributions will span methodological developments and clinical applications, reflecting current challenges and opportunities in the field.

Skill Level: Intermediate

		563-03-001. Physics-Informed Multi-Parametric Machine Learning Model to Differentiate Glioblastomas and Brain Metastases Seyyed Ali Hosseini, Archith Rajan, Nesrine Rahmouni, Arthur C. Macedo, Tevy Chan, Gleb Bezgin, Steven Brem, Stephen Bagley, Mohamad Nazem-Zadeh, Pedro Rosa‐Neto, Suyash Mohan, Sanjeev Chawla McGill University, Montreal, Canada Impact: By incorporating physics laws into multiparametric MRI, machine learning based models can better capture tumor biology, leading to development of more transparent and reliable clinical decision tools.
		563-03-002. A Comparison of Radio-Pathomic, Diffusion, and Perfusion Imaging Features for Identifying Pseudoprogression in Gliomas Samuel Bobholz, Aleksandra Winiarz, Benjamin Chao, Hope Reecher, Daniel Kim, Allison Lowman, Biprojit Nath, Savannah Duenweg, Adam Lahrache, Mrina Mtenga, Fitzgerald Kyereme, Michael Barrett, Jennifer Connelly, Elaine Tanhehco, Max Krucoff, Mohit Agarwal, Jamie Jacobsohn, Rupen Desai, Jennifer Tuscher, Peter LaViolette Medical College of Wisconsin, Wauwatosa, United States of America Impact: Radio-pathomic maps outperform current advanced imaging techniques in distinguishing true progression from pseudoprogression in post-treatment glioma patients in the UCSF-PTGBM dataset, potentially offering clinicians a more accurate, non-invasive tool for treatment decisions without requiring additional contrast agents or scan time.
		563-03-003. Microstructure characterisation of brain tumours by VERDICT MRI: a benchmark of deep learning methods Matteo Figini, Zheng Yu, Marco Palombo, Michele Bailo, Antonella Castellano, Daniel Alexander, Eleftheria Panagiotaki University College London, London, United Kingdom Impact: This benchmark establishes performance baselines and a reproducible evaluation framework for Deep-Learning-based VERDICT, laying groundwork for further work for the clinical translation of fast non-invasive brain tumour microstructure imaging.
		563-03-004. Metabolic, Perfusion, and Diffusion Imaging Enhance Diagnosis and Prognosis of H3K27-Altered Diffuse Midline Gliomas QIU JUN, Jinyuan Weng, Wei Wei The First Affiliated Hospital of USTC, Hefei, China Impact: This study demonstrates that integrating APTw, ASL, and DKI metrics significantly improves diagnosis and prognosis for H3K27-altered DMGs, enabling better clinical decision-making and providing a non-invasive imaging biomarker for patient stratification.
		563-03-005. Optimized Time-Dependent Diffusion MRI for Preoperative Molecular Subtyping of Adult Diffuse Gliomas Xin Ge, Yuhui Xiong, Jing Zhang, Wen Wang Tangdu Hospital, Fourth Military Medical University, Xi'an, China Impact: Optimized TDD-MRI with Bayesian IMPULSED produces reproducible microstructural maps and accurate preoperative IDH/1p/19q subtyping, informing surgical strategy and trial-ready therapy selection, while enabling voxelwise prognostication, treatment monitoring, and multicenter standardization of diffusion-time biomarkers.
		563-03-006. Impact of incorporating T2 relaxation in the VERDICT model for brain tumor microstructure imaging Lukas Lundholm, Oscar Jalnefjord, Mikael Montelius, Mats Laesser, Thomas Bontell, Alba Corell, Asgeir Jakola, Isabella Björkman-Burtscher, Maria Ljungberg University of Gothenburg, Gothenburg, Sweden Impact: Including compartment-specific T2 relaxation in the VERDICT model substantially alters parameter estimates in brain tumors, indicating that neglecting T2 differences may bias results and that accounting for this effect may be critical for accurate estimation of VERDICT-derived parameters.
		563-03-007. Characterisation of brain tumour type using multi-parametric information from GE-SE EPIK Fabian Küppers, Keith George Ciantar, Seong Dae Yun, Gabriele Stoffels, Christian Filß, Norbert Galldiks, Felix Mottaghy, M. Eline Kooi, Karl-Josef Langen, Philipp Lohmann, N. Jon Shah Forschungszentrum Juelich, Juelich, Germany Impact: GE-SE EPIK-derived mean OEF and R₂’ change with respect to different FET TBR thresholds used for tumour VOI segmentation, while R₂’ show potential to differentiate astrocytoma. Parameter combinations improve the significance of tumour type differences compared to single parameters.
		563-03-008. Accurate and Robust Brain Tumor Classification with Vision Transformer Models Dr. Abdullah Asiri Najran University, Najran, Saudi Arabia Impact: This study advances clinical neuroimaging by demonstrating that Swin Transformer models significantly enhance brain tumor classification accuracy.The findings empower radiologists with reliable diagnostic support tools,encourage broader application of transformer-based AI in medical imaging, and inspire future research on cross-modality generalization.
		563-03-009. Comparative Evaluation of Brain Tissue Stiffness in Pituitary Adenoma and Healthy Volunteers Using MR Elastography Shivani Tripathi, Priyanka Bhat, Ashish Suri, Ajay Garg, Yogesh k Mariappan, Sandeep Ganji, Poonam Choudhary, S Senthil Kumaran All India Institute of Medical Sciences, New Delhi, India Impact: Brain MR Elastography non-invasively quantifies pituitary adenoma stiffness, aiding preoperative planning and potentially improving surgical precision and patient outcomes.
		563-03-010. MRI-Driven Finite Element Analysis for Patient-Specific Glioma Biomechanics Enamul Bhuiyan, Md Mazder Rahman, Xiaohong Joe Zhou University of Illinois Chicago, Chicago, United States of America Impact: This physics-informed pipeline enables the estimation of interpretable biomechanical properties, offering a scalable, noninvasive, and clinically translatable tool for glioma assessment with further validations.
		563-03-011. Hybrid 3D U-Net and Hyperdimensional Architecture for Efficient Brain Tumor Analysis Muhammad Saad Ullah, Iqra Akhtar COMSATS University Islamabad, Islamabad, Pakistan Impact: Through this novel coil technology, a fundamental hardware technology provides us higher SNR and data acquisition acceleration across MRI applications. This empowers the clinical diagnostics with improved resolution and speed, and offers researchers to ground-break new biomedical studies.
		563-03-012. Sodium MRI and MR Elastography as Quantitative Biomarkers in Pediatric Brain Tumors: A Translational Approach L Tyler Williams, Katrina Milbocker, Sabrina Vander Wiele, Grace McIlvain, Em Triolo, N Ece Yilmaz, Suraj Serai, Raisa Amiruddin, Peter Madsen, Lina Chihoub, Curtis Johnson, Timothy Roberts, Aashim Bhatia Children's Hospital of Philadelphia, Philadelphia, United States of America Impact: This translational pilot study helps establish the value of sodium (²³Na) MRI and MR Elastography (MRE) as imaging biomarkers in pediatric brain tumors which may improve neurosurgical planning and our ability to detect treatment resistant tumor regions and tumor progression.
		563-03-013. NOE MRI for Improved Brain Tumor Infiltration Blake Benyard, Narayan Datt Soni, Anshuman Swain, Nishi Srivastava, Sunil Khokhar, Nadir Yehya, Yi Fan, Dushyant Kumar, RAVI PRAKASH REDDY NANGA, Michael Frenneaux, Mohammad Haris, Ravinder Reddy University of Pennsylvania, Philadelphia, United States of America Impact: This method could enhance the assessment of tumor boundaries for surgical and radiation treatment planning.
		563-03-014. Imaging Microstructural Environment in Patients with Brain Metastasis and Meningioma: Preliminary Findings from SANDI Sang-Young Kim, Eunju Kim, Jinwoo Hwang, Eung Yeop Kim, Sung Tae Kim, Beomseok Sohn Philips Korea, Seoul, Korea, Republic of Impact: Diffusion MRI with SANDI model would offer a non-invasive way to distinguish brain tumor phenotypes by probing cell body and neurite features, which may provide microstructural biomarkers.