Related Experiment Video
Updated: Nov 8, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.1K
A new deep learning method for blood vessel segmentation in retinal images based on convolutional kernels and
Manuel E Gegundez-Arias1, Diego Marin-Santos1, Isaac Perez-Borrero1
1Vision, Prediction, Optimisation and Control Systems Department, Science and Technology Research Centre, University of Huelva, Avenida de las Fuerzas Armadas s/n, 21007, Huelva, Spain.
Computer Methods and Programs in Biomedicine
|April 21, 2021
Summary
This study introduces a deep learning method for segmenting retinal blood vessels in eye fundus images. The novel approach achieves superior accuracy compared to existing methods, offering a promising tool for clinical applications.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Automatic monitoring of retinal blood vessels is crucial for diagnosing ocular vascular diseases.
- Accurate segmentation of the vascular tree in fundus images aids in clinical assessment.
Purpose of the Study:
- To present an efficient and accurate deep learning-based method for retinal blood vessel segmentation.
- To develop a robust tool for analyzing vascular anomalies in eye fundus images.
Main Methods:
- A convolutional neural network (CNN) based on a simplified U-Net architecture was employed.
- The network utilized residual blocks and batch normalization for improved performance.
- A novel loss function incorporating pixel distance to the vascular tree was implemented for training.
Main Results:
- The method was evaluated on the DRIVE, STARE, and CHASE_Db1 databases.
- It demonstrated robust performance across diverse fundus image origins.
- Achieved accuracy comparable to or exceeding state-of-the-art methods.
Conclusions:
- The proposed deep learning method offers superior performance for retinal blood vessel segmentation.
- It is a promising candidate for integration into clinical diagnostic tools.
