Related Experiment Video
Updated: Jul 19, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
570
Multi-level spatial-temporal and attentional information deep fusion network for retinal vessel segmentation.
1School of Information Science and Technology, Southwest Jiaotong University, 611756, Chengdu, People's Republic of China.
Physics in Medicine and Biology
|August 11, 2023
Summary
This study introduces MSAFNet, a novel deep learning network for accurate retinal vessel segmentation in fundus images. The method significantly improves automated diagnosis by enhancing detection capabilities.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Accurate segmentation of retinal vessels is crucial for intelligent disease diagnosis systems.
- Existing methods face challenges in achieving precise and robust vessel segmentation in fundus images.
Purpose of the Study:
- To develop a robust and effective method for segmenting retinal blood vessels in human color fundus images.
- To enhance the performance and robustness of automated retinal vessel segmentation.
Main Methods:
- A novel multi-level spatial-temporal and attentional information deep fusion network (MSAFNet) was developed.
- The network integrates multi-level spatial-temporal encoding and a Self-Attention module.
- An encoder-decoder structure fuses features for final segmentation.
Main Results:
- MSAFNet demonstrated superior performance compared to state-of-the-art methods on four public datasets.
- Achieved high Accuracy (up to 96.96%) and Area Under Curve (up to 98.78%) scores.
- Achieved high Specificity (up to 99.08%) on DRIVE and STARE datasets.
Conclusions:
- The MSAFNet method exhibits strong learning and representation capabilities for retinal vessel detection.
- The accurate detection of retinal blood vessels can serve as a valuable tool for assisting in medical diagnosis.
