Related Experiment Video
Updated: Aug 13, 2025
12:28
Quantification of Vascular Parameters in Whole Mount Retinas of Mice with Non-Proliferative and Proliferative Retinopathies
Published on: March 12, 2022
3.7K
Segmentation of retinal vessels based on MRANet
Sanli Yi1, Yanrong Wei1, Gang Zhang1
1School of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650500, Yunnan, China.
Heliyon
|January 23, 2023
Summary
This study introduces a novel Multi-Scale Residual Attention Network (MRANet) for improved retinal vessel segmentation. The MRANet enhances diagnostic accuracy by effectively segmenting fine capillaries and reducing noise interference.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Retinal vessel segmentation is vital for diagnosing eye diseases.
- Challenges include segmenting fine capillaries and handling noise.
Purpose of the Study:
- To develop an improved method for retinal vessel segmentation.
- To address limitations of existing methods in capillary segmentation and noise resistance.
Main Methods:
- Proposed a Multi-Scale Residual Attention Network (MRANet) based on U-Net.
- Introduced Multi-Level Feature Fusion (MLF) block for effective information collection.
- Utilized attention blocks for feature weighting and DropBlock layer to reduce parameters and overfitting.
Main Results:
- Achieved high accuracy (0.9698 on DRIVE, 0.9755 on CHASE_DB1) and AUC (0.9899 on DRIVE, 0.9893 on CHASE_DB1).
- Demonstrated superior segmentation compared to other methods.
- Ensured continuity and completeness in blood vessel segmentation.
Conclusions:
- MRANet significantly improves retinal vessel segmentation accuracy and robustness.
- The proposed network effectively segments fine vessels and reduces noise.
- This method holds promise for enhanced computer-aided diagnosis of eye disorders.
