AI-Based Analysis in MR Imaging
Digital Poster
Analysis Methods
Thursday, 14 May 2026
Machine Learning and Artificial Intelligence changing MR image processing, quantitative imaging and workflow optimisation.
Skill Level: Basic,Intermediate,Advanced
|
661-04-001.
Cross-modal Brain Connectivity Prediction with Topology-aware Signed Graph Diffusion Model
Impact: Our topology-aware signed graph diffusion model enhances the prediction of structural connectivity from functional connectivity, enhancing more reliable clinical evaluations in neurological diagnostics and addressing a key need in areas where such technologies are crucial.
|
||
|
661-04-002.
Cross-compartment phenotyping and machine-learning classification of pelvic foor dysfunction using MR defecography
Impact: The proposed framework provides a standardized way to quantify pelvic floor dysfunction and identify clinically meaningful phenotypes, improving triage between physiotherapy and surgical intervention. It may reduce misclassification, support treatment planning, and guide future multicenter outcome validation.
|
||
|
661-04-003.
Prediction of estimated risk for subsequent motor disorders in infants with external hydrocephalus using machine learning and
Impact: This study provides a practical tool for early risk assessment in infants with EH, aiding clinical decision-making. The integrated multi-omics findings offer new insights into the potential neurobiological mechanisms associated with motor delay in this population.
|
||
|
661-04-004.
Dynamic modeling of MRI-based tumor habitats during radiotherapy in a murine glioma model
Impact: This study used four biology-based mathematical models to track temporal dynamics of tumor habitats in a murine glioma model, revealing habitat-specific radiotherapy responses and providing a predictive framework to guide personalized, more effective treatment strategies
|
||
|
661-04-005.
Artificial intelligence-enhanced MRI interpretation of tumor invasion: A novel tumor–MRF invasion score for rectal cancer
Impact: TMIS overcomes the
limitations of conventional MRI-based evaluation, providing a more streamlined
and integrated approach for prognostic risk stratification in rectal cancer
patients.
|
||
|
661-04-006.
3D-CNN-Based In Situ pSAR Prediction of Implanted Pedicle Screw Systems under 1.5 T MRI Across Multiple Human Models
Impact: Current implant RF-safety assessment lack patient
specificity or clinical scalability. This deep-learning surrogate enables rapid
in-situ pSAR prediction from anatomical and B₁⁺ field
inputs, supporting anatomically diverse cases and providing clinicians with a
practical tool for individualized MRI safety assessment.
|
||
|
661-04-007.
MRI-based multimodal model to predict lymph node metastasis after neoadjuvant chemoradiotherapy in rectal cancer
Impact: This study presents an interpretable
MRI-based multimodal model that accurately predicts lymph node metastasis after
neoadjuvant chemoradiotherapy, providing a noninvasive tool to guide
individualized surgical strategies and improve clinical decision-making for
locally advanced rectal cancer.
|
||
|
661-04-008.
Fast and robust dictionary generation for multiparametric cardiac mapping with variable timing using a transformer network
Impact: Transformer networks enable rapid, precise, and
robust generation of cardiac T1-T2 mapping dictionaries with variable
acquisition timing, replacing 30-minute simulations with 10-second inference
and allowing on-scanner dictionary generation for multiparametric mapping in
clinical workflows.
|
||
|
661-04-009.
AI assistance enhances radiologists' accuracy in rectal MRI T-stage : A retrospective multi-reader, multi-case study
Impact: The provision of AI-generated
visual results led to improved radiologist accuracy of T-stage assessment,
particularly enabling less experienced radiologists to increase their
diagnostic consistency and effectively narrow the gap with their more
experienced radiologists.
|
||
|
661-04-010.
Detecting Metabolic Heterogeneity within Leukodystrophy Patients using Unsupervised CEST Z-Spectral Analysis
Impact: Enables rapid, reproducible, voxel-wise metabolic screening to highlight white matter pathology and guide segmentation and disease progression analysis without ground truth labels.
|
||
|
661-04-011.
Placental patch and whole-Uterus dual-branch U-Mamba-based Architecture for Chorioamnionitis detection from functional MRI
Impact: This is a proof-of-concept for the use of functional MRI combined with state-of-the-art deep learning for in-vivo detection of chorioamnionitis. Early identification could enable timely intervention and contribute to reducing adverse neonatal outcomes such as cerebral palsy and necrotising enterocolitis.
|
||
|
661-04-012.
AI-Driven Workflow Optimization and Energy Efficiency in MRI: Toward Sustainable Imaging in Resource-Limited Settings
Impact: This study demonstrates that integrating AI-driven acceleration and workflow optimization in MRI can significantly reduce energy use and costs, enabling sustainable imaging practices. It encourages broader adoption in resource-limited settings and prompts further research on AI-enabled, carbon-efficient radiology workflows.
|
||
|
661-04-013.
Learning-Based Synthetic MRI Post-Processing Framework for Automated Contrast Optimization and Brain Segmentation
Impact: This framework enables data-driven optimization of MRI contrast synthesis within a physically interpretable model using Synthetic MRI multi-parametric maps, thereby achieving higher downstream performance of multi-contrast generation and clinically meaningful subcortical delineation critical for Huntington’s disease (HD) staging and monitoring.
|