Diffusion MRI Reconstruction Methods
Digital Poster
Diffusion
Monday, 11 May 2026
This session will deal with diffusion MRI reconstruction methods
|
365-06-001.
Energy-based Profile Encoding for 3D Multi-slab Diffusion-weighted imaging (EPEN)
Impact: EPEN enables flexible 3D-multislab dMRI acquisition without oversampling or restrictive sampling patterns. By combining a sampling strategy that avoids structured artifacts and a MAP-optimization employing CNN-based energy-priors, it robustly suppresses slab-boundary artifacts while preserving resolution and SNR-efficiency for clinical imaging.
|
||
|
365-06-002.
Quantitative Evaluation of Deep Learning–Based Diffusion MRI Reconstruction in Neuro MRI
Impact: Deep learning reconstruction enables faster diffusion MRI with preserved ADC accuracy, improving workflow efficiency and patient comfort while maintaining quantitative reliability for neuroimaging and therapy monitoring.
|
||
|
365-06-003.
Radial Basis Function Network for Highly Accelerated Radial Diffusion Spectrum Imaging (DSI)
Impact: The proposed reconstruction enables synthesis of missing q-space samples from up to four-fold undersampled data, reducing RDSI acquisition time to about four minutes. Trained directly on the acquired data, it offers a robust, versatile, and scanner-independent framework.
|
||
|
365-06-004.
Anatomically Guided Reconstruction for Multi-Echo Stimulated Echo (MESTIM) Radial Diffusion Spectrum Imaging (RDSI)
Impact: The proposed pipeline substantially enhances SNR and image quality in high b-value diffusion imaging. Our results indicate that high-resolution interleaved b=0 images could improve the spatial resolution of MESTIM EPI data, supporting broader clinical use and improved microstructural mapping.
|
||
|
365-06-005.
Don't Miss a Beat: Pulsatile Motion Matters in ULF DWI
Impact: Without accurate motion state characterisation, ultra-low-field multi-shot 3D DWI can suffer from signal dropout from nonrigid motion. The many-shot paradigm of ULF-DWI can be retrospectively sorted by cardiac phase (systole/diastole), markedly reducing motion artifacts.
|
||
|
365-06-006.
AI-Based Complex Averaging and Image Enhancement for Diffusion-Weighted Imaging
Impact: A complex data processing incorporating image
averaging and enhancement using artificial intelligence models enables considerable
gains in signal-to-noise ratio, spatial resolution, and scan time in
single-shot echo-planar diffusion-weighted imaging.
|
||
|
365-06-007.
Enhanced Voxel-Level fODF Reconstruction from Single-Shell dMRI Using V-NET with Adaptive Multi-Scale Attention
Impact: The proposed V-NET with adaptive multi-scale attention efficiently reconstructs voxel-level fiber orientation distribution functions (fODFs) from clinically practical single-shell diffusion MRI data. It enables fast and reliable tractography for routine clinical and surgical use.
|
||
|
365-06-008.
Model once, Map all - Moving forward in Diffusion MRI
Impact: Our method delivers a complete microstructure suite in one run, cutting runtime and pipeline complexity. It is flexible to protocol, robust across sites, species, and acquisition types.
|
||
|
365-06-009.
Evaluation of small FOV DWI with complex signal averaging and deep learning reconstruction for clinically robust DWI at 3T
Impact: This study
demonstrated the feasibility of combining multiple novel DWI technologies to
acquire DWI images with limited FOVs, without aliasing artifacts, reduced
geometrical distortions, and enhanced SNR in clinical settings, as a crucial
step toward clinically robust diffusion imaging.
|
||
|
365-06-010.
Clinical Applications of 5.0T Artificial Intelligence Deep Learning Reconstruction Diffusion-Weighted Imaging in Cranial MRI
Impact: First to validate 5.0T DWI Deep Learning Reconstruction for clinical use. It offers neuro-oncology ultra-fast, ultra-high-resolution imaging, significantly improving brain tumor diagnostic accuracy and driving 5T MR clinical translation.
|
||
|
365-06-011.
Deep Learning Reconstruction Effects in High b-Value DWI: From Fruit Phantoms to Breast Tissue
Impact: This
proof-of-concept study, confirms the feasibility of using fruits as low-cost phantoms
to be tested under breast MRI acquisition parameters. DL-rs-EPI allows for higher image quality at high
b-values, supporting its integration into breast DWI-based protocols for
improving lesions’ detection.
|
||
|
365-06-012.
Scanner-Adaptive Coil-Level Denoising for Diffusion MRI Using Explicit Noise Priors
Impact: We propose a novel deep learning based method to perform coil-specific and scanner-adaptive denoising of low-SNR diffusion MRI data to enhance structural details and boost the signal-to-noise ratio.
|
||
|
365-06-013.
0.55T Prostate Diffusion-Weighted Imaging Using Multi-Shot EPI and Self-Supervised Learning Reconstruction
Impact: This study supplies a reliable and accurate prostate DWI method at 0.55T, leveraging multi-shot EPI acquisition and self-supervised unrolled joint reconstruction. This method will allow for large cohort prostate patient studies at 0.55T.
|
||
|
365-06-014.
Self-Supervised Denoising of Pancreatic DWI for Cancer Assessment Using Physics-Informed Implicit Neural Representations
Impact: We present a self-supervised, physically constrained INR framework that denoises pancreatic DWI while preserving diffusion behavior. By enhancing image fidelity and diffusion-derived biomarkers, particularly sADC and DDC, our method offers substantial potential for more accurate and reliable pancreatic MRI diagnostics.
|
||
|
365-06-015.
Deep learning-based phase correction and denoising improves clinical multi-shot DWI
Impact: DL-based
phase correction and denoising can improve msDWI of the abdomen in clinical exams,
which still suffer from suboptimal SNR and motion-induced artifacts.
|