Physics-Driven Deep Learning for Computational Magnetic Resonance Imaging: Combining physics and machine learning for improved medical imaging

被引:0
|
作者
Hammernik, Kerstin [1 ,2 ]
Kustner, Thomas [3 ,4 ]
Yaman, Burhaneddin [5 ]
Huang, Zhengnan [6 ]
Rueckert, Daniel [7 ,8 ]
Knoll, Florian [9 ,10 ]
Akcakaya, Mehmet [11 ,12 ]
机构
[1] Tech Univ Munich, Dept Informat, D-85748 Garching, Germany
[2] Imperial Coll London, Dept Comp, London SW7 2AZ, England
[3] Kings Coll London, Sch Biomed Engn & Imaging Sci, London, England
[4] Univ Hosp Tubingen, Grp Med Imaging & Data Anal, D-72076 Tubingen, Germany
[5] Univ Minnesota, Elect Engn, Minneapolis, MN 55455 USA
[6] New York Univ, Ctr Biomed Imaging, Sch Med, New York, NY 10016 USA
[7] Tech Univ Munich, Artificial Intelligence Med & Healthcare, D-85748 Garching, Germany
[8] Imperial Coll London, Visual Informat Proc, Dept Comp, London SW7 2AZ, England
[9] New York Univ, Radiol, Ctr Biomed Imaging, Grossman Sch Med, New York, NY USA
[10] Friedrich Alexander Univ Erlangen Nuremberg, Dept Artificial Intelligence Biomed Engn, D-91052 Erlangen, Germany
[11] Harvard Med Sch, Boston, MA USA
[12] Univ Minnesota, Minneapolis, MN 55455 USA
基金
英国工程与自然科学研究理事会; 美国国家科学基金会; 美国国家卫生研究院;
关键词
Deep learning; Inverse problems; Magnetic resonance imaging; Computational modeling; Pipelines; Signal processing; Task analysis; K-SPACE; RECONSTRUCTION; NETWORKS; MODELS;
D O I
10.1109/MSP.2022.3215288
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
Physics-driven deep learning methods have emerged as a powerful tool for computational magnetic resonance imaging (MRI) problems, pushing reconstruction performance to new limits. This article provides an overview of the recent developments in incorporating physics information into learning-based MRI reconstruction. We consider inverse problems with both linear and nonlinear forward models for computational MRI and review the classical approaches for solving these. We then focus on physics-driven deep learning approaches, covering physics-driven loss functions, plug-and-play (PnP) methods, generative models, and unrolled networks. We highlight domain-specific challenges, such as real- and complex-valued building blocks of neural networks, and translational applications in MRI with linear and nonlinear forward models. Finally, we discuss common issues and open challenges, and we draw connections to the importance of physics-driven learning when combined with other downstream tasks in the medical imaging pipeline.
引用
收藏
页码:98 / 114
页数:17
相关论文
empty
未找到相关数据