A generic fundus image enhancement network boosted by frequency self-supervised representation learning

Heng Li; Haofeng Liu; Huazhu Fu; Yanwu Xu; Hai Shu; Ke Niu; Yan Hu; Jiang Liu

doi:10.1016/j.media.2023.102945

A generic fundus image enhancement network boosted by frequency self-supervised representation learning

Med Image Anal. 2023 Dec:90:102945. doi: 10.1016/j.media.2023.102945. Epub 2023 Sep 9.

Authors

Heng Li¹, Haofeng Liu², Huazhu Fu³, Yanwu Xu⁴, Hai Shu⁵, Ke Niu⁶, Yan Hu⁷, Jiang Liu⁸

Affiliations

¹ Research Institute of Trustworthy Autonomous Systems and Department of Computer Science and Engineering, Southern University of Science and Technology, Shenzhen, China. Electronic address: lih3@sustech.edu.cn.
² Research Institute of Trustworthy Autonomous Systems and Department of Computer Science and Engineering, Southern University of Science and Technology, Shenzhen, China. Electronic address: liuhf2020@mail.sustech.edu.cn.
³ Institute of High Performance Computing (IHPC), Agency for Science, Technology and Research (A*STAR), Singapore.
⁴ School of Future Technology, South China University of Technology, Guangzhou, China; Pazhou Lab, Guangzhou, China.
⁵ Department of Biostatistics, School of Global Public Health, New York University, NY, USA.
⁶ Computer School, Beijing Information Science and Technology University, Beijing, China.
⁷ Research Institute of Trustworthy Autonomous Systems and Department of Computer Science and Engineering, Southern University of Science and Technology, Shenzhen, China. Electronic address: huy3@sustech.edu.cn.
⁸ Research Institute of Trustworthy Autonomous Systems and Department of Computer Science and Engineering, Southern University of Science and Technology, Shenzhen, China; Department of Computer Science and Engineering, Southern University of Science and Technology, Shenzhen, China; Guangdong Provincial Key Laboratory of Brain-inspired Intelligent Computation, Southern University of Science and Technology, Shenzhen, China. Electronic address: liuj@sustech.edu.cn.

PMID: 37703674
DOI: 10.1016/j.media.2023.102945

Abstract

Fundus photography is prone to suffer from image quality degradation that impacts clinical examination performed by ophthalmologists or intelligent systems. Though enhancement algorithms have been developed to promote fundus observation on degraded images, high data demands and limited applicability hinder their clinical deployment. To circumvent this bottleneck, a generic fundus image enhancement network (GFE-Net) is developed in this study to robustly correct unknown fundus images without supervised or extra data. Levering image frequency information, self-supervised representation learning is conducted to learn robust structure-aware representations from degraded images. Then with a seamless architecture that couples representation learning and image enhancement, GFE-Net can accurately correct fundus images and meanwhile preserve retinal structures. Comprehensive experiments are implemented to demonstrate the effectiveness and advantages of GFE-Net. Compared with state-of-the-art algorithms, GFE-Net achieves superior performance in data dependency, enhancement performance, deployment efficiency, and scale generalizability. Follow-up fundus image analysis is also facilitated by GFE-Net, whose modules are respectively verified to be effective for image enhancement.

Keywords: Fundus image enhancement; Seamless coupling; Self-supervised representation learning; Structure-aware representations.