Related Experiment Video
Updated: Dec 12, 2025
15:48
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
22.9K
pFISTA-SENSE-ResNet for parallel MRI reconstruction.
Tieyuan Lu1, Xinlin Zhang1, Yihui Huang1
1Department of Electronic Science, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, School of Electronic Science and Engineering, Xiamen University, Xiamen 361005, China.
Journal of Magnetic Resonance (San Diego, Calif. : 1997)
|August 8, 2020
Summary
This study introduces a novel deep learning network for faster, high-quality parallel magnetic resonance imaging reconstruction. The proposed method improves image reconstruction accuracy and robustness compared to existing techniques.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Image Reconstruction
Background:
- Magnetic resonance imaging (MRI) is crucial for clinical diagnosis but suffers from long acquisition times.
- Accelerating MRI acquisition using sparse sampling and parallel imaging presents computational challenges for image reconstruction.
- Current deep learning methods offer promising reconstruction but lack interpretability.
Purpose of the Study:
- To develop a novel deep learning network for high-quality parallel MRI reconstruction.
- To address the challenge of fast and accurate image reconstruction in accelerated MRI.
- To improve the interpretability of deep learning models in MRI reconstruction.
Main Methods:
- Designed a novel network structure inspired by sparse iterative reconstruction principles.
- Integrated a residual structure to enhance the network's performance.
- Evaluated the proposed network on a public knee MRI dataset.
Main Results:
- The proposed network achieved lower reconstruction error compared to state-of-the-art methods.
- Demonstrated superior robustness across various undersampling patterns.
- Outperformed both deep learning-based and optimization-based reconstruction techniques.
Conclusions:
- The novel residual-enhanced sparse iterative network enables high-quality parallel MRI reconstruction.
- The method offers a promising solution for accelerating MRI acquisition while maintaining image fidelity.
- This approach enhances accuracy and robustness in accelerated MRI, paving the way for faster clinical diagnosis.
