Related Experiment Video
Updated: Jul 31, 2025
Retinal Vascular Reactivity as Assessed by Optical Coherence Tomography Angiography
Published on: March 26, 2020
A lightweight network guided with differential matched filtering for retinal vessel segmentation
Yubo Tan1, Shi-Xuan Zhao1, Kai-Fu Yang1
1The MOE Key Laboratory for Neuroinformation, Radiation Oncology Key Laboratory of Sichuan Province, University of Electronic Science and Technology of China, China.
Insights
This study introduces a new deep learning model for precise retinal vessel segmentation in fundus images, improving detection of thin vessels and reducing errors in challenging areas.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Retinal vessel morphology indicates cardiovascular health, making fundus image analysis crucial.
- Automated retinal vessel segmentation faces challenges with thin vessels, lesions, and low contrast.
Purpose of the Study:
- To develop an advanced deep learning model for accurate thin vessel segmentation in fundus images.
- To address limitations of existing methods regarding vessel breakage and false positives in complex regions.
Main Methods:
- Proposed a novel network, differential matched filtering guided attention UNet (DMF-AU).
- Incorporated differential matched filtering for initial vessel identification, feature anisotropic attention, and a multiscale consistency constrained backbone.
- Utilized these components to enhance learning of vascular details and spatial linearity.
Main Results:
- The DMF-AU model demonstrated superior performance in thin vessel segmentation compared to existing algorithms.
- Achieved high accuracy on specially designed criteria for vessel segmentation tasks.
- The model proved effective in handling areas with lesions and low contrast.
Conclusions:
- DMF-AU is a high-performance, lightweight model for precise retinal vessel segmentation.
- The proposed architecture effectively addresses challenges in segmenting thin vessels and reduces false positives.
- This work offers a valuable tool for ophthalmologists analyzing fundus images for cardiovascular health assessment.
Abstract:
The geometric morphology of retinal vessels reflects the state of cardiovascular health, and fundus images are important reference materials for ophthalmologists. Great progress has been made in automated vessel segmentation, but few studies have focused on thin vessel breakage and false-positives in areas with lesions or low contrast. In this work, we propose a new network, differential matched filtering guided attention UNet (DMF-AU), to address these issues, incorporating a differential matched filtering layer, feature anisotropic attention, and a multiscale consistency constrained backbone to perform thin vessel segmentation. The differential matched filtering is used for the early identification of locally linear vessels, and the resulting rough vessel map guides the backbone to learn vascular details. Feature anisotropic attention reinforces the vessel features of spatial linearity at each stage of the model. Multiscale constraints reduce the loss of vessel information while pooling within large receptive fields. In tests on multiple classical datasets, the proposed model performed well compared with other algorithms on several specially designed criteria for vessel segmentation. DMF-AU is a high-performance, lightweight vessel segmentation model. The source code is at https://github.com/tyb311/DMF-AU.
