Image synthesis in contrast MRI based on super resolution reconstruction with multi-refinement cycle-consistent generative adversarial networks

Kun Wu, Yan Qiang (), Kai Song, Xueting Ren, WenKai Yang, Wanjun Zhang, Akbar Hussain and Yanfen Cui
Additional contact information
Kun Wu: Taiyuan University of Technology
Yan Qiang: Taiyuan University of Technology
Kai Song: Taiyuan University of Technology
Xueting Ren: Taiyuan University of Technology
WenKai Yang: Taiyuan University of Technology
Wanjun Zhang: Beijing Institution of Technology
Akbar Hussain: Taiyuan University of Technology
Yanfen Cui: Shanxi Medical University

Journal of Intelligent Manufacturing, 2020, vol. 31, issue 5, No 10, 1215-1228

Abstract: Abstract In the field of medical image processing represented by magnetic resonance imaging (MRI), synthesizing the complementary target contrast of the target patient from the existing contrast has obvious medical significance for assisting doctors in making clinical diagnoses. To satisfy the image translation problem between different MRI contrasts (T1 and T2), a generative adversarial network is proposed that works in an end-to-end manner at image level. The low-frequency and high-frequency information of the image is preserved by using multi-stage optimization learning aided by adversarial loss, the loss of perceptual consistency and the loss of cyclic consistency, as it results in preserving the same contrast anatomical structure of the source domain supervisely when the perceptual pixel distribution of the target contrast is learned perfectly. To integrate different penalties (L1 and L2) organically, adaptive weights are set for the error sensitivity of the penalty function in the present total loss function, the aim being to achieve adaptive optimization of each stage of generating high-resolution images. In addition, a new net structure called multi-skip connection residual net is proposed to refine medical image details step by step with multi-stage optimization. Compared with the existing technology, the present method is more advanced. The contrast conversion of T1 and T2 in MRI is validated, which can help to shorten the imaging time, improve the imaging quality, and effectively assist doctors with diagnoses.

Keywords: Synthesis; Contrast MRI; Generative adversarial network; Multi-stage; Cyclic consistency; Multi-skip (search for similar items in EconPapers)
Date: 2020
References: View complete reference list from CitEc
Citations:

Downloads: (external link)
http://link.springer.com/10.1007/s10845-019-01507-7 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:joinma:v:31:y:2020:i:5:d:10.1007_s10845-019-01507-7

Ordering information: This journal article can be ordered from
http://www.springer.com/journal/10845

DOI: 10.1007/s10845-019-01507-7

Access Statistics for this article

Journal of Intelligent Manufacturing is currently edited by Andrew Kusiak

More articles in Journal of Intelligent Manufacturing from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().