A ROI based self-supervised strategy for retinal image analysis

Shiyun Dong; Qiang Chen

doi:10.1117/12.2667274

27 January 2023 A ROI based self-supervised strategy for retinal image analysis

Shiyun Dong, Qiang Chen

Proceedings Volume 12550, International Conference on Optical and Photonic Engineering (icOPEN 2022); 125501I (2023) https://doi.org/10.1117/12.2667274
Event: International Conference on Optical and Photonic Engineering (icOPEN 2022), 2022, ONLINE, China

Abstract

With the maturity of deep learning methods and the collection of a large number of retinal images, many deep learning models have been used for retinopathy analysis, but most of them are supervised methods. That is, the training of the models relies heavily on a large number of annotated retinal images, which is sometimes difficult to satisfy in medical fields including ophthalmology. This paper explores the application of self-supervised methods that do not require a large amount of labeled data in the analysis of retinal images. Firstly, we propose a novel self-supervised task of column dislocation and compare the performance of various self-supervised tasks on the retinopathy classification, verifying our proposed task’s effectiveness. Then, by considering the characteristics of retinal images, namely the Region Of Interest (ROI) in Spectral Domain Optical Coherence Tomography (SD-OCT) images is less than half of the whole images, we also propose a ROI based self-supervised strategy to improve the feature representation ability. The experiments on two public datasets for retinal image classification and segmentation demonstrate that the proposed strategy is effective compared with several existed self-supervised learning methods.

Citation Download Citation

Shiyun Dong and Qiang Chen "A ROI based self-supervised strategy for retinal image analysis", Proc. SPIE 12550, International Conference on Optical and Photonic Engineering (icOPEN 2022), 125501I (27 January 2023); https://doi.org/10.1117/12.2667274

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available