Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Simultaneous q-Space Sampling Optimization and Reconstruction for Fast and High-fidelity Diffusion Magnetic Resonance Imaging

Jing Yang, Jian Cheng, Cheng Li, Wenxin Fan, Juan Zou, Ruoyou Wu, Shanshan Wang

TL;DR

This work tackles the challenge of long dMRI acquisition times by proposing SSOR, a framework that simultaneously optimizes q-space sampling and reconstructs high angular resolution data. It leverages a continuous spherical-harmonics representation to model q-space signals, learns sparse sampling directions, and uses an end-to-end neural network with $l_1$ sparsity and total-variation regularization to recover the full dMRI dataset. The approach demonstrates superior quantitative performance (e.g., PSNR and SSIM) and qualitative fidelity across acceleration factors and exhibits robustness to noise and out-of-distribution diffusion weights. With its flexible sampling strategy and strong performance on HCP data, SSOR holds promise for faster, clinically practical diffusion MRI workflows.

Abstract

Diffusion Magnetic Resonance Imaging (dMRI) plays a crucial role in the noninvasive investigation of tissue microstructural properties and structural connectivity in the \textit{in vivo} human brain. However, to effectively capture the intricate characteristics of water diffusion at various directions and scales, it is important to employ comprehensive q-space sampling. Unfortunately, this requirement leads to long scan times, limiting the clinical applicability of dMRI. To address this challenge, we propose SSOR, a Simultaneous q-Space sampling Optimization and Reconstruction framework. We jointly optimize a subset of q-space samples using a continuous representation of spherical harmonic functions and a reconstruction network. Additionally, we integrate the unique properties of diffusion magnetic resonance imaging (dMRI) in both the q-space and image domains by applying $l1$-norm and total-variation regularization. The experiments conducted on HCP data demonstrate that SSOR has promising strengths both quantitatively and qualitatively and exhibits robustness to noise.

Simultaneous q-Space Sampling Optimization and Reconstruction for Fast and High-fidelity Diffusion Magnetic Resonance Imaging

TL;DR

This work tackles the challenge of long dMRI acquisition times by proposing SSOR, a framework that simultaneously optimizes q-space sampling and reconstructs high angular resolution data. It leverages a continuous spherical-harmonics representation to model q-space signals, learns sparse sampling directions, and uses an end-to-end neural network with sparsity and total-variation regularization to recover the full dMRI dataset. The approach demonstrates superior quantitative performance (e.g., PSNR and SSIM) and qualitative fidelity across acceleration factors and exhibits robustness to noise and out-of-distribution diffusion weights. With its flexible sampling strategy and strong performance on HCP data, SSOR holds promise for faster, clinically practical diffusion MRI workflows.

Abstract

Diffusion Magnetic Resonance Imaging (dMRI) plays a crucial role in the noninvasive investigation of tissue microstructural properties and structural connectivity in the \textit{in vivo} human brain. However, to effectively capture the intricate characteristics of water diffusion at various directions and scales, it is important to employ comprehensive q-space sampling. Unfortunately, this requirement leads to long scan times, limiting the clinical applicability of dMRI. To address this challenge, we propose SSOR, a Simultaneous q-Space sampling Optimization and Reconstruction framework. We jointly optimize a subset of q-space samples using a continuous representation of spherical harmonic functions and a reconstruction network. Additionally, we integrate the unique properties of diffusion magnetic resonance imaging (dMRI) in both the q-space and image domains by applying -norm and total-variation regularization. The experiments conducted on HCP data demonstrate that SSOR has promising strengths both quantitatively and qualitatively and exhibits robustness to noise.
Paper Structure (13 sections, 7 equations, 2 figures, 3 tables)

This paper contains 13 sections, 7 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Method
  3. q-Space Sampling Optimization block
  4. Reconstruction block
  5. Training Loss
  6. Experiments
  7. Datasets
  8. Implementation Details
  9. Results
  10. Experimental results on in-distribution data
  11. Experimental results on out-of-distribution data
  12. Conclusions
  13. Acknowledgments

Figures (2)

  • Figure 1: The illustration of SSOR framework.
  • Figure 2: Qualitative results of different methods.