|
| 引用本文: | 程 航,蔡 昕,姜小平,杨 光,贾守强,聂生东.基于深度学习的欠采样磁共振图像重建研究综述[J].软件工程,2023,26(6):1-5.【点击复制】 |
|
| |
|
|
|
|
| 摘 要: 针对加速磁共振成像(Magnetic Resonance Imaging, MRI)中,传统算法对压缩感知欠采样磁共振(Magnetic Resonance, MR)图像重建质量欠佳的问题,以基于深度学习的欠采样MR图像重建算法为研究对象,系统性地概述图像重建算法中模型驱动与数据驱动方法的工作原理,分析各自的特性与优点、缺点,对重建方法进行对比讨论,并列举了部分算法在数据集上的表现。结果表明,大部分算法在数据集上的结构相似性指标(Structure Similarity Index Measure, SSIM)为0.87—0.96。依据重建方法现存的不足与当前的研究趋势,提出3种MR图像重建算法未来的发展方向。 |
| 关键词: MRI;欠采样图像重建;深度学习;模型驱动;数据驱动 |
|
中图分类号: TP391.4
文献标识码: A
|
|
| Overview of Under-sampled Magnetic Resonance Image Reconstruction Based on Deep Learning |
|
CHENG Hang1, CAI Xin1, JIANG Xiaoping2, YANG Guang3, JIA Shouqiang4, NIE Shengdong1
|
(1.School of Health Science and Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China;
2.Shanghai Kangda Colorful Medical Technology Co., Ltd., Shanghai 200444, China;
3. School of Physics and Electronic Science, East China Normal University, Shanghai 200062, China;
4.Department of Imaging, Jinan People's Hospital Affiliated to Shandong First Medical University, Jinan 271100, China)
2909994580@qq.com; cxsmic@163.com; xp_jiang@msn.com; gyang@phy.ecnu.edu.cn; jshqlw@163.com; nsd4647@163.com
|
| Abstract: Traditional algorithms perform poorly on reconstructing compressed sensing under-sampled Magnetic Resonance (MR) images in accelerated Magnetic Resonance Imaging (MRI). In order to solve this problem, taking under-sampled MR Image reconstruction algorithm based on deep learning as the research object, this paper proposes to systematically summarize the working principles of model-driven and data-driven methods in image reconstruction algorithms, and analyze their characteristics, advantages and disadvantages. The reconstruction methods are compared and discussed, and performance of some algorithms on the dataset is listed. The results show that most algorithms have a Structure Similarity Index Measure (SSIM) of 0.87-0.96 on the dataset. The future development directions of the three MR image reconstruction algorithms are proposed based on the existing shortcomings of the reconstruction methods and current research trends. |
| Keywords: MRI; under-sampled image reconstruction; deep learning; model-driven; data-driven |
|
|
|
|
