Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Joint Edge Optimization Deep Unfolding Network for Accelerated MRI Reconstruction

Yue Cai, Yu Luo, Jie Ling, Shun Yao

TL;DR

This work tackles the prolonged scan time in MRI by incorporating edge priors through a joint edge optimization model. It introduces a non-edge probability map $P_{ne}$ and co-regularizes it with the image, all embedded in a deep unfolding framework with ERN and IDN modules to learn priors automatically. The method demonstrates superior PSNR gains across multi-coil and single-coil datasets under various undersampling schemes, validating the edge-guided reconstruction and the efficacy of stage-wise, non-shared unfolding. The findings highlight the practical value of explicit edge-information utilization and learned priors for accelerated MRI, with potential extension to other low-level vision tasks.

Abstract

Magnetic Resonance Imaging (MRI) is a widely used imaging technique, however it has the limitation of long scanning time. Though previous model-based and learning-based MRI reconstruction methods have shown promising performance, most of them have not fully utilized the edge prior of MR images, and there is still much room for improvement. In this paper, we build a joint edge optimization model that not only incorporates individual regularizers specific to both the MR image and the edges, but also enforces a co-regularizer to effectively establish a stronger correlation between them. Specifically, the edge information is defined through a non-edge probability map to guide the image reconstruction during the optimization process. Meanwhile, the regularizers pertaining to images and edges are incorporated into a deep unfolding network to automatically learn their respective inherent a-priori information.Numerical experiments, consisting of multi-coil and single-coil MRI data with different sampling schemes at a variety of sampling factors, demonstrate that the proposed method outperforms other compared methods.

Joint Edge Optimization Deep Unfolding Network for Accelerated MRI Reconstruction

TL;DR

This work tackles the prolonged scan time in MRI by incorporating edge priors through a joint edge optimization model. It introduces a non-edge probability map and co-regularizes it with the image, all embedded in a deep unfolding framework with ERN and IDN modules to learn priors automatically. The method demonstrates superior PSNR gains across multi-coil and single-coil datasets under various undersampling schemes, validating the edge-guided reconstruction and the efficacy of stage-wise, non-shared unfolding. The findings highlight the practical value of explicit edge-information utilization and learned priors for accelerated MRI, with potential extension to other low-level vision tasks.

Abstract

Magnetic Resonance Imaging (MRI) is a widely used imaging technique, however it has the limitation of long scanning time. Though previous model-based and learning-based MRI reconstruction methods have shown promising performance, most of them have not fully utilized the edge prior of MR images, and there is still much room for improvement. In this paper, we build a joint edge optimization model that not only incorporates individual regularizers specific to both the MR image and the edges, but also enforces a co-regularizer to effectively establish a stronger correlation between them. Specifically, the edge information is defined through a non-edge probability map to guide the image reconstruction during the optimization process. Meanwhile, the regularizers pertaining to images and edges are incorporated into a deep unfolding network to automatically learn their respective inherent a-priori information.Numerical experiments, consisting of multi-coil and single-coil MRI data with different sampling schemes at a variety of sampling factors, demonstrate that the proposed method outperforms other compared methods.
Paper Structure (20 sections, 21 equations, 7 figures, 3 tables, 1 algorithm)

This paper contains 20 sections, 21 equations, 7 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related work
  3. Accelerated parallel imaging
  4. Advances in Deep Unfolding for MRI reconstruction
  5. Method
  6. Model
  7. Algorithms
  8. Edge Information Solver
  9. Reconstructed Image Solver
  10. Deep Unfolding Network
  11. Training Strategy
  12. Experimental Evaluation
  13. Implementation Details
  14. Dataset
  15. Quantitative and Qualitative Evaluation
  16. ...and 5 more sections

Figures (7)

  • Figure 1: (a) The low-quality MR image; (b) The edges of (a); (c) The high-quality MR image; (d) The edges of (c).
  • Figure 2: The overall iterative framework of the joint edge optimization deep unfolding network for accelerated MRI
  • Figure 3: Non-edge probability map visualization. (a) and (f) is the initialization and ground truth of non-edge probability map respectively. From (b) to (e), we visualize the non-edge probability map of increasing stages.
  • Figure 4: Results on multi-coil dataset using 6x random sampling. (a) Ground truth and random mask with the accelerated factor of 6. (b) The zero-filled reconstruction result and the error map; (c)-(i) The reconstructed images and the error maps by U-NetHyun_Kim_Lee_Lee_Seo_2018, DCCNNschlemper2017deep, MoDLaggarwal2018modl, VS-Netduan2019vs, RecurrentVarNetyiasemis2022recurrent, EAMRIyang2023fast and our proposed method respectively.
  • Figure 5: Results on single-coil dataset using 6x random Cartesian sampling. (a) Ground truth and random Cartesian mask with the accelerated factor of 6. (b) The zero-filled reconstruction result and the error map; (c)-(i) The reconstructed images and the error maps by U-NetHyun_Kim_Lee_Lee_Seo_2018, ADMM-CSNetyang2018admm, DCCNNschlemper2017deep, RefineGANquan2018compressed, HQS-Netxin2022learned, EAMRIyang2023fast and our proposed method respectively.
  • ...and 2 more figures