Related Experiment Video
Updated: Nov 1, 2025
12:28
Quantification of Vascular Parameters in Whole Mount Retinas of Mice with Non-Proliferative and Proliferative Retinopathies
Published on: March 12, 2022
3.9K
A high resolution representation network with multi-path scale for retinal vessel segmentation
Zefang Lin1, Jianping Huang1, Yingyin Chen1
1Zhuhai Interventional Medical Center, Zhuhai Precision Medical Center, Zhuhai People's Hospital, Zhuhai Hospital Affiliated with Jinan University, Jinan University, Zhuhai, Guangdong 519000, PR China.
Computer Methods and Programs in Biomedicine
|June 19, 2021
Summary
A novel Multi-Path Scale Network (MPS-Net) improves automatic retinal vessel segmentation (RVS) in fundus images. This method enhances early diagnosis of eye diseases by accurately detecting retinal blood vessels.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Accurate automatic retinal vessel segmentation (RVS) is crucial for early diagnosis of ophthalmologic diseases.
- Challenges in RVS include vascular complexities, lesions, and optic disc edges in fundus images.
Purpose of the Study:
- To propose a novel high-resolution representation network with multi-path scale (MPS-Net) for improved RVS.
- To enhance the extraction of retinal blood vessels in fundus images.
Main Methods:
- Developed MPS-Net featuring a high-resolution main path and lower-resolution branch paths with multi-path scale modules.
- Introduced a hard-focused cross-entropy loss function to guide the network towards learning challenging features.
Main Results:
- Evaluated MPS-Net on DRIVE, STARE, CHASE, and synthetic datasets.
- Demonstrated superior performance over existing methods in F1-score, sensitivity, G-mean, and Matthews correlation coefficient.
Conclusions:
- MPS-Net shows promising segmentation performance for retinal vessel analysis.
- The method has potential for real-world applications and adaptation to other medical imaging tasks.
