A Structure-Consistency GAN for Unpaired AS-OCT Image Inpainting

Bai, Guanhua1; Li, Sanqian2; Zhang, He1; Higashita, Risa1,2,3; Liu, Jiang1,2; Li, Jie1; Zhang, Meng1

10.1007/978-3-031-44013-7_15

2023

10th International Workshop on Ophthalmic Medical Image Analysis (OMIA)

0302-9743

1611-3349

978-3-031-44012-0

OPHTHALMIC MEDICAL IMAGE ANALYSIS, OMIA 2023

14096

OCT 12, 2023

null,Vancouver,CANADA

GEWERBESTRASSE 11, CHAM, CH-6330, SWITZERLAND

SPRINGER INTERNATIONAL PUBLISHING AG

Anterior segment optical coherence tomography (AS-OCT) is a crucial imaging modality in ophthalmology, providing valuable insights into corneal pathologies. However, during AS-OCT imaging, intense signals in highly reflective regions can easily lead to saturation effects, resulting in pronounced stripes across the cornea. It compromises the image visual quality and impacts automated ophthalmic analysis. To address this issue, we propose an unsupervised Structure-Consistency Generative Adversarial Network (SC-GAN) that captures the underlying semantic structural knowledge in both the spatial domain and frequency space within the generative model. This strategy aims to mitigate the influence of bright stripes and restore corneal structural details in AS-OCT images. Specifically, SC-GAN introduces a stripe perceptual loss to extract visual representations by utilizing the perceptual similarity between striped and stripe-free images. Moreover, Fourier feature mapping is adopted to learn high-frequency information, thereby achieving crucial structure consistency. The experimental results demonstrate that the proposed SC-GAN can removes stripes while preserving crucial corneal structures, surpassing the competing algorithms. Furthermore, we validate the benefits of SC-GAN in the corneal segmentation task.

AS-OCT Inpainting GAN structural consistency

其他

英语

EI ; CPCI-S

Computer Science ; Engineering ; Ophthalmology

Computer Science, Artificial Intelligence ; Computer Science, Theory & Methods ; Engineering, Biomedical ; Ophthalmology

WOS:001116062200015

Web of Science

1.Changchun University, Changchun, China
2.Research Institute of Trustworthy Autonomous Systems and Department of Computer Science and Engineering, Southern University of Science and Technology, Shenzhen, China
3.Tomey Corporation, Nagoya, Japan

Bai, Guanhua,Li, Sanqian,Zhang, He,et al. A Structure-Consistency GAN for Unpaired AS-OCT Image Inpainting[C]. GEWERBESTRASSE 11, CHAM, CH-6330, SWITZERLAND:SPRINGER INTERNATIONAL PUBLISHING AG,2023.