Post-Processing of Cardiovascular MRI
Digital Poster
Cardiovascular
Monday, 11 May 2026
This session will showcase emerging post-processing techniques in cardiovascular MRI, including the identification of new biomarkers, AI-based denoising, motion correction and reconstruction.
Skill Level: Basic,Intermediate,Advanced
|
365-02-001.
CMR-derived biventricular models and biomarker assessment to characterize fitness-related cardiac geometries
Impact: This study
integrates an automated extraction of biventricular geometries from CMR with cardiopulmonary
exercise tests and body composition scans to investigate healthy cardiac geometrical
patterns associated with physical activity
|
||
|
365-02-002.
Generation of CFD-enhanced synthetic 5D Flow MRI
Impact: Synthetic 5D flow MRI datasets with known ground truth enable controlled
evaluation of free-breathing acquisition and reconstruction strategies. Such
data provide a benchmark for optimizing motion compensation and developing both
learning-based and physics-based methods.
|
||
|
365-02-003.
Age-Related Changes in Cardiac Anatomy Using Cardiac Magnetic Resonance-Derived Geometries in a Healthy Swiss Cohort
Impact: This open-source dataset and automated pipeline enable
standardized extraction of cardiac shape models. By quantifying geometric
remodeling across age, these models provide insights into cardiac anatomy and
establish a basis for personalized assessment of cardiac structure and
function.
|
||
|
365-02-004.
One-Shot Multi-Tissue Inversion Time Prediction for Cardiac MRI
Impact: We present a single-shot, multi-tissue TI prediction method for LGE imaging using parametric inter-frame distance maps and classification networks, validated on diverse multi-site data, enabling robust, automated TI nulling across protocols, contrast agents, patient conditions, and artifacts.
|
||
|
365-02-005.
Native T1 Mapping Derived-Radiomics Model for Predicting Left Ventricular Adverse Remodeling in STEMI Patients After PCI
Impact: This study showed the value of non-contrast enhanced T1 mapping in the early identification of high-risk left ventricular adverse remodeling after PCI in STEMI patients, which provides valuable information for the identification of patients with high morbidity and mortality risk.
|
||
|
365-02-006.
One-click AI-assisted analysis of joint bright- and black-blood LGE and T2 mapping in acute STEMI patients
Impact: Artificial intelligence-driven analysis of SPOT-MAPPING allows for a
faster and easier diagnosis, while avoiding intra- and inter-observer
variability, therefore decreasing CMR image analysis complexity.
|
||
|
365-02-007.
Decoding Myocardial Heterogeneity: A CMR Habitat Analysis Framework for MACE Prediction after STEMI
Impact: This study innovatively integrates CMR radiomics with habitat analysis, providing a novel imaging-based framework for post-STEMI risk stratification and clinical decision support to improve individualized patient management and outcomes.
|
||
|
365-02-008.
Denoising DENSE MRI with Deep Learning for Accurate Myocardial Strain Quantification
Impact: This work presents a deep
learning-based DENSE denoiser that enables recovery of high-quality strain
information from low-SNR DENSE images. This approach could expand use of DENSE
at lower field strengths where SNR limitations currently reduce clinical
utility.
|
||
|
365-02-009.
Rigid Motion Estimation using Accelerated Coordinate Descent (REACT) for MR Imaging
Impact: This study establishes the feasibility of the coordinate descent approach for autofocus motion correction, providing an efficient and computationally viable alternative to gradient-based methods. The proposed method is applicable to various acquisition schemes, including both Cartesian and non-Cartesian sampling.
|
||
|
365-02-010.
The application value of deep learning reconstruction in optimizing MRI myocardial delayed enhancement imaging
Impact: Deep
learning reconstructed late gadolinium enhancement images showed enhanced
visualization of myocardial scar and fibrosis, which may contribute to an
increased detection rate of lesions, particularly those that are early and
small, thereby facilitating more accurate diagnoses.
|
||
|
365-02-011.
Assessing Carotid Plaque Vulnerability Without Gadolinium: A Physics and Mask-Guided Deep Learning Approach
Impact: This gadolinium-free synthesis network PM-GAN(Physics and Mask-Guided GAN) provides a reliable alternative for plaque vulnerability assessment, potentially changing clinical protocols and benefiting patients contraindicated for contrast agents.
|
||
|
365-02-012.
Deep learning-based motion compensated reconstruction for self-gated cardiac MRA utilizing self-supervised finetuning
Impact: High-resolution MRA acquisition with short
and predictable scan times is achieved. Resulting image quality is comparable
to the clinical reference. The proposed method strengthens the potential of MRA
as a clinically viable, non-invasive alternative to computed tomography
angiography.
|
||
|
365-02-013.
Radiomic Feature Selection Strategies for Differentiating Fabry Disease from Hypertrophic Cardiomyopathy on Cardiac MRI
Impact: Differentiating
FD from HCM based on cine radiomics benefits from a pairing strategy that
integrates filtered features, strict redundancy suppression, and appropriate
selectors. MI is relatively robust across settings, whereas Chi2 benefits most
from curated wavelet subbands plus stricter correlation.
|
||
|
365-02-014.
Predicting cardiac magnetic resonance image based survival with machine learning
Impact: The historical registry data was shown to have prognostic value but is non-linear in nature. The Random Survival Forests benefited from the added historical data and improved prognostic value which is important for improving cardiac risk prediction.
|