Dual-tier segmentation approach for hard exudates in fundus images with U-net convolutional neural network

Soumika Guduru; Emma Munch; Veda Murthy; Brent J. Liu

doi:10.1117/12.2612925

4 April 2022 Dual-tier segmentation approach for hard exudates in fundus images with U-net convolutional neural network

Soumika Guduru, Emma Munch, Veda Murthy, Brent J. Liu

Proceedings Volume PC12037, Medical Imaging 2022: Imaging Informatics for Healthcare, Research, and Applications; PC120370A (2022) https://doi.org/10.1117/12.2612925
Event: SPIE Medical Imaging, 2022, San Diego, California, United States

Abstract

In this paper we demonstrate an automated pipeline for segmentation of the optic disk and hard exudates in retinal fundus images. The image is preprocessed to normalize luminosity and contrast throughout the image, and blood vessels are removed to reduce interference with the high-contrast properties of the optic disk and hard exudates. A pretrained VGG-16 U-net convolutional neural network is used for segmentation of the optic disk and ROI definition for hard exudates. The exudates are then segmented with a filtering method to isolate a rough segmentation, which is used to identify inputs to a U-net pretrained on exudate regions.

Conference Presentation

Citation Download Citation

Soumika Guduru, Emma Munch, Veda Murthy, and Brent J. Liu "Dual-tier segmentation approach for hard exudates in fundus images with U-net convolutional neural network", Proc. SPIE PC12037, Medical Imaging 2022: Imaging Informatics for Healthcare, Research, and Applications, PC120370A (4 April 2022); https://doi.org/10.1117/12.2612925

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