Accelerating cDTI with deep learning-based tensor de-noising and breath hold reduction. a step towards improved efficiency and clinical feasibility
File(s)
Published version
Author(s)
Type
Journal Article
Abstract
Background
Cardiac Diffusion Tensor Imaging (cDTI) non-invasively provides unique insights into cardiac microstructure. Current protocols require multiple breath-hold repetitions to achieve adequate signal-to-noise ratio, resulting in lengthy scan times. The aim of this study was to develop a cDTI de-noising method that would enable the reduction of repetitions while preserving image quality.
Methods
We present a novel de-noising framework for cDTI acceleration centred on three fundamental advances: (1) a paradigm shift from image-based to tensor-space de-noising that better preserves structural information, (2) an ensemble of Vision Transformer-based models specifically optimised for tensor processing through adversarial training, and (3) a sophisticated data augmentation strategy that maximises training data utilisation through dynamic repetition selection.
Results
Our approach reduces scan times by a factor of up to 4 while achieving a 20% reduction in cDTI maps errors over existing de-noising methods (Table 1) and preserving anatomical features such as infarct characterisation and transmural cardiomyocyte orientation patterns. Crucially, our proposed method succeeds in clinical cases where other algorithms previously failed.
Conclusions
This demonstrates substantial improvements in cDTI acquisition efficiency, achieving up to 4-fold scan time reduction (3-5 breath-holds) while maintaining diagnostic accuracy across diverse cardiac pathologies.
Cardiac Diffusion Tensor Imaging (cDTI) non-invasively provides unique insights into cardiac microstructure. Current protocols require multiple breath-hold repetitions to achieve adequate signal-to-noise ratio, resulting in lengthy scan times. The aim of this study was to develop a cDTI de-noising method that would enable the reduction of repetitions while preserving image quality.
Methods
We present a novel de-noising framework for cDTI acceleration centred on three fundamental advances: (1) a paradigm shift from image-based to tensor-space de-noising that better preserves structural information, (2) an ensemble of Vision Transformer-based models specifically optimised for tensor processing through adversarial training, and (3) a sophisticated data augmentation strategy that maximises training data utilisation through dynamic repetition selection.
Results
Our approach reduces scan times by a factor of up to 4 while achieving a 20% reduction in cDTI maps errors over existing de-noising methods (Table 1) and preserving anatomical features such as infarct characterisation and transmural cardiomyocyte orientation patterns. Crucially, our proposed method succeeds in clinical cases where other algorithms previously failed.
Conclusions
This demonstrates substantial improvements in cDTI acquisition efficiency, achieving up to 4-fold scan time reduction (3-5 breath-holds) while maintaining diagnostic accuracy across diverse cardiac pathologies.
Date Issued
2025-12-11
Date Acceptance
2025-10-07
Citation
Journal of Cardiovascular Magnetic Resonance, 2025, 27 (2)
ISSN
1097-6647
Publisher
Elsevier BV
Journal / Book Title
Journal of Cardiovascular Magnetic Resonance
Volume
27
Issue
2
Copyright Statement
© 2025 The Author(s). Published by Elsevier Inc. on behalf of Society for Cardiovascular Magnetic Resonance.
License URL
Identifier
10.1016/j.jocmr.2025.101971
Subjects
Cardiac MRI Deep learning Diffusion tensor imaging MRI DWI, Diffusion Weighted Image
E2A, Sheetlet Angle
EPI, Echo-Planar Imaging
FA, Fractional Anisotropy
HA, Helix Angle
LGE, Late Gadolinium Enhancement
LLS, Linearized Least Squares
LV, Left Ventricle
MAAE, Mean Absolute Angle Error
MAE, Mean Absolute Error
MD, Mean Diffusivity
MI, Myocardial Infarction
SIT, Situs Inversus Totalis
SNR, Signal to Noise Ratio
STEAM, STimulated Echo Acquisition Mode
WGUF, Wasserstein GAN UFormer
MRI, magnetic resonance imaging
CT, computed tomography
GAN, generative adversarial network
EPI, echo-planar imaging
SENSE, sensitivity encoding
GRAPPA, generalized autocalibrating partially parallel acquisition
BH, breath hold
CN, channel normalization
Publication Status
Published
Article Number
101971
Date Publish Online
2025-11-08