Skip to content

Iterative Feature Refinement Network for Compressed Sensing MRI

Notifications You must be signed in to change notification settings

PooyanRezaeipour/IFR-Net-Code

 
 

Repository files navigation

IFR-Net-Code

IFR-Net: Iterative Feature Refinement Network for Compressed Sensing MRI

About The Code

The Code based on the method described in the following paper:
IFR-Net: Iterative Feature Refinement Network for Compressed Sensing MRI
Author: Yiling Liu, Qiegen Liu, Minghui Zhang, Qingxin Yang, Shanshan Wang and Dong Liang
IEEE Trans. Comput. Imag., vol. 6, pp. 434-446, 2020.
Version : 4.0
The code and the algorithm are for non-comercial use only.
Copyright 2019, Department of Electronic Information Engineering, Nanchang University.

Abstract

To improve the compressive sensing MRI (CS-MRI) approaches in terms of fine structure loss under high acceleration factors, we have proposed an iterative feature refinement model (IFR-CS), equipped with fixed transforms, to restore the meaningful structures and details. Nevertheless, the proposed IFR-CS still has some limitations, such as the selection of hyper-parameters, a lengthy reconstruction time, and the fixed sparsifying transform. To alleviate these issues, we unroll the iterative feature refinement procedures in IFR-CS to a supervised model-driven network, dubbed IFR-Net. Equipped with training data pairs, both regularization parameter and the utmost feature refinement operator in IFR-CS become trainable. Additionally, inspired by the powerful representation capability of convolutional neural network (CNN), CNN-based inversion blocks are explored in the sparsity-promoting denoising module to generalize the sparsity-enforcing operator. Extensive experiments on both simulated and in vivo MR datasets have shown that the proposed network possesses a strong capability to capture image details and preserve well the structural information with fast reconstruction speed.

Overall structure of the IFR-Net


Result


Fig. 1 Real-valued reconstruction results on brain image. Sampling pattern:10% pseudo radial sampling. Left to right: Ground truth, IFR-CS, ADMM-Net, and IFR-NET.


Fig. 2 Complex-valued reconstruction results on brain image. Sampling pattern:25% 2D random sampling. Left to right: Ground truth, IFR-CS, ADMM-Net and IFR-NET.

The link for some of compared methods

D5C5 https://github.com/js3611/Deep-MRI-Reconstruction
ADMM-Net https://github.com/yangyan92/Deep-ADMM-Net

Other Related Projects

  • Variable augmented neural network for decolorization and multi-exposure fusion [Paper] [Code] [Slide]

  • VST-Net: Variance-stabilizing transformation inspired network for Poisson denoising [Paper] [Code]

  • Iterative scheme-inspired network for impulse noise removal [Paper] [Code]

  • A Comparative Study of CNN-based Super-resolution Methods in MRI Reconstruction and Its Beyond [Paper] [Code]

  • Progressively distribution-based Rician noise removal for magnetic resonance imaging [Paper1] [Paper2] [Code]

  • Complex-valued MRI data from SIAT [Data]

About

Iterative Feature Refinement Network for Compressed Sensing MRI

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Cuda 45.6%
  • MATLAB 29.4%
  • C++ 15.9%
  • Python 5.0%
  • Shell 2.0%
  • C 1.3%
  • Other 0.8%