Related Experiment Video
Updated: Sep 29, 2025
07:23
Retinal Vascular Reactivity as Assessed by Optical Coherence Tomography Angiography
Published on: March 26, 2020
7.6K
Improving sensitivity and connectivity of retinal vessel segmentation via error discrimination network
Guoye Lin1, Hanhua Bai1, Jie Zhao1,2
1Guangdong Provincial Key Laboratory of Medical Image Processing, Guangdong Province Engineering Laboratory for Medical Imaging and Diagnostic Technology, School of Biomedical Engineering, Southern Medical University, Guangzhou, Guangdong, China.
Medical Physics
|March 26, 2022
Summary
This study introduces a novel deep learning approach for retinal vessel segmentation, improving accuracy for thin vessels and connectivity. The method enhances early diagnosis of eye diseases by providing more robust segmentation results.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Automated retinal vessel segmentation is vital for diagnosing eye conditions.
- Deep learning methods excel but struggle with thin vessels and connectivity.
- Existing approaches lack focus on segmentation accuracy for challenging cases.
Purpose of the Study:
- To develop an improved automated retinal vessel segmentation method.
- To enhance segmentation accuracy for thin vessels and vessel connectivity.
- To address limitations in current deep learning-based segmentation techniques.
Main Methods:
- A novel approach using an error discrimination network (D) alongside a segmentation network (S).
- Network D identifies incorrect vessel pixel predictions from network S.
- Network S is trained to minimize errors flagged by D, incorporating artificial thin vessel samples for enhanced sensitivity.
Main Results:
- Achieved state-of-the-art sensitivity and G-Mean scores on STARE, DRIVE, and HRF datasets.
- Demonstrated top performance in topology-relevant metrics (Conn, Inf, Cor), indicating excellent vessel connectivity.
- Maintained competitive F1-score and AUC, with AUC ranking first on the STARE dataset.
Conclusions:
- The proposed method offers an accurate and robust solution for challenging retinal vessel segmentation.
- The error discrimination approach effectively improves segmentation of difficult vessels.
- This technique holds promise for advancing early diagnosis and treatment of ophthalmological diseases.
