Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

3D Anatomical Structure-guided Deep Learning for Accurate Diffusion Microstructure Imaging

Xinrui Ma, Jian Cheng, Wenxin Fan, Ruoyou Wu, Yongquan Ye, Shanshan Wang

TL;DR

Diffusion MRI provides non-invasive brain microstructure estimates but traditional model fitting requires extensive gradient sampling, limiting clinical use. The paper proposes an anatomy-guided deep learning framework that leverages macro-level anatomical priors and cross-parameter information to reconstruct multi-model diffusion maps from sparsely sampled q-space, enabling faster acquisitions while preserving accuracy. Findings on public HCP data show PSNR 30.51 ± 0.58 and SSIM 0.97 ± 0.004 with about 15× acceleration over dense sampling, demonstrating strong performance in reconstructing DTI and NODDI maps. Significance lies in improved acquisition efficiency and boundary fidelity, enabling more feasible clinical diffusion microstructure imaging through an anatomy-informed, multi-model joint estimation approach.

Abstract

Diffusion magnetic resonance imaging (dMRI) is a crucial non-invasive technique for exploring the microstructure of the living human brain. Traditional hand-crafted and model-based tissue microstructure reconstruction methods often require extensive diffusion gradient sampling, which can be time-consuming and limits the clinical applicability of tissue microstructure information. Recent advances in deep learning have shown promise in microstructure estimation; however, accurately estimating tissue microstructure from clinically feasible dMRI scans remains challenging without appropriate constraints. This paper introduces a novel framework that achieves high-fidelity and rapid diffusion microstructure imaging by simultaneously leveraging anatomical information from macro-level priors and mutual information across parameters. This approach enhances time efficiency while maintaining accuracy in microstructure estimation. Experimental results demonstrate that our method outperforms four state-of-the-art techniques, achieving a peak signal-to-noise ratio (PSNR) of 30.51$\pm$0.58 and a structural similarity index measure (SSIM) of 0.97$\pm$0.004 in estimating parametric maps of multiple diffusion models. Notably, our method achieves a 15$\times$ acceleration compared to the dense sampling approach, which typically utilizes 270 diffusion gradients.

3D Anatomical Structure-guided Deep Learning for Accurate Diffusion Microstructure Imaging

TL;DR

Diffusion MRI provides non-invasive brain microstructure estimates but traditional model fitting requires extensive gradient sampling, limiting clinical use. The paper proposes an anatomy-guided deep learning framework that leverages macro-level anatomical priors and cross-parameter information to reconstruct multi-model diffusion maps from sparsely sampled q-space, enabling faster acquisitions while preserving accuracy. Findings on public HCP data show PSNR 30.51 ± 0.58 and SSIM 0.97 ± 0.004 with about 15× acceleration over dense sampling, demonstrating strong performance in reconstructing DTI and NODDI maps. Significance lies in improved acquisition efficiency and boundary fidelity, enabling more feasible clinical diffusion microstructure imaging through an anatomy-informed, multi-model joint estimation approach.

Abstract

Diffusion magnetic resonance imaging (dMRI) is a crucial non-invasive technique for exploring the microstructure of the living human brain. Traditional hand-crafted and model-based tissue microstructure reconstruction methods often require extensive diffusion gradient sampling, which can be time-consuming and limits the clinical applicability of tissue microstructure information. Recent advances in deep learning have shown promise in microstructure estimation; however, accurately estimating tissue microstructure from clinically feasible dMRI scans remains challenging without appropriate constraints. This paper introduces a novel framework that achieves high-fidelity and rapid diffusion microstructure imaging by simultaneously leveraging anatomical information from macro-level priors and mutual information across parameters. This approach enhances time efficiency while maintaining accuracy in microstructure estimation. Experimental results demonstrate that our method outperforms four state-of-the-art techniques, achieving a peak signal-to-noise ratio (PSNR) of 30.510.58 and a structural similarity index measure (SSIM) of 0.970.004 in estimating parametric maps of multiple diffusion models. Notably, our method achieves a 15 acceleration compared to the dense sampling approach, which typically utilizes 270 diffusion gradients.

Paper Structure

This paper contains 14 sections, 2 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Task Formulation
  4. Anatomical prior module
  5. Loss function
  6. EXPERIMENTS
  7. Dataset
  8. Experimental Settings
  9. Results
  10. Comparison of state-of-the-art methods
  11. Ablation Study: Comparison of Anatomical Priors
  12. CONCLUSION
  13. Compliance with ethical standards
  14. Acknowledgments

Figures (2)

  • Figure 1: An overview of the proposed framework. The input consists of two components: the upper branch with six DWIs and a b0 image, and the lower branch with tissue probability maps from T1w extraction. These inputs are concatenated and processed by the network to yield DTI and NODDI microstructural parameters, including fractional anisotropy(FA), mean diffusivity(MD), axial diffusivity(AD), intracellular volume fraction(Vic), isotropic volume fraction(Viso) and orientation dispersion index(OD).
  • Figure 2: The ground truth, estimated parametric maps and corresponding error maps based on MF, q_DL, CNN, MESC-SD, and Ours in a test subject with 6 diffusion directions per shell at b-values of 1000, 2000, and 3000 $s/mm^{2}$.