Related Experiment Video
Updated: Jan 9, 2026
08:51
Magnetic Resonance Imaging of Multiple Sclerosis at 7.0 Tesla
Published on: February 19, 2021
9.7K
Frequency-spatial synergistic network for accelerated multi-contrast MRI reconstruction
Ke Li1, Dong Liang2, Guoqing Chen1
1School of Mathematical Science, Inner Mongolia University, Hohhot, 010000, Inner Mongolia, China; Inner Mongolia Key Laboratory of Mathematical Modeling and Scientific Computing, Hohhot, 010000, Inner Mongolia, China.
Summary
The new Multi-Contrast Dual-Domain (MCDD) network accelerates Magnetic Resonance Imaging (MRI) reconstruction by effectively integrating multi-contrast information. This approach significantly improves image quality and reduces artifacts in accelerated MRI scans.
Area of Science:
- Medical Imaging
- Biomedical Engineering
- Computer Vision
Background:
- Accelerating Magnetic Resonance Imaging (MRI) is crucial for clinical efficiency.
- Multi-contrast MRI reconstruction utilizes auxiliary modalities to enhance target contrast reconstruction.
- Key challenges include capturing cross-contrast global dependencies and integrating complementary information.
Purpose of the Study:
- To propose a novel network, the Multi-Contrast Dual-Domain (MCDD) network, for accelerated multi-contrast MRI reconstruction.
- To effectively capture global dependencies and integrate complementary information across different MRI contrasts.
- To improve the quality and efficiency of accelerated MRI scans.
Main Methods:
- The proposed Multi-Contrast Dual-Domain (MCDD) network leverages frequency-domain information for global dependency capture.
- It employs a Single-Contrast Learning (SCL) module as a prior to guide a Multi-Contrast Learning (MCL) module.
- Both modules extract features using Frequency (Fre) and Spatial (Spa) blocks, fused by an Adaptive Fusion Mechanism (AFM).
Main Results:
- MCDD demonstrated superior performance over state-of-the-art methods on BraTS, MRBrainS, and fastMRI datasets.
- Achieved significant Peak Signal-to-Noise Ratio (PSNR) improvements: 2.42 dB (4×), 2.11 dB (8×), 1.63 dB (16×), and 0.82 dB (32×).
- Qualitative results showed enhanced structural fidelity and reduced artifacts in reconstructed MRI images.
Conclusions:
- The MCDD network is an effective solution for accelerated MRI reconstruction.
- It successfully addresses the challenges of cross-contrast dependency and information integration.
- The method offers improved image quality and artifact reduction for faster MRI acquisition.
