Related Experiment Video
Updated: Jul 13, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
565
A generative adversarial neural network with multi-attention feature extraction for fundus lesion segmentation
Haiying Yuan1, Mengfan Dai2, Cheng Shi2
1Faculty of Information Technology, Beijing University of Technology, No.100 Pingleyuan, Chaoyang District, Beijing, 100124, People's Republic of China. yhycn@126.com.
International Ophthalmology
|October 18, 2023
Summary
A new generative adversarial network accurately segments diabetic retinopathy lesions in fundus images. This AI approach aids in precise disease detection and treatment planning for diabetic eye conditions.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Diabetic retinopathy diagnosis relies on segmenting lesions in fundus images.
- Accurate segmentation is challenging due to lesion characteristics like scattered distribution and feature similarity.
Purpose of the Study:
- To develop an advanced generative adversarial network for precise diabetic retinopathy segmentation.
- To improve the extraction of local and global features for better lesion identification.
Main Methods:
- A generative adversarial network incorporating multi-attention feature extraction was designed.
- An improved U-Net with self-attention and external attention mechanisms was used for feature extraction.
- A PatchGAN-based discriminative network enhanced segmentation accuracy.
Main Results:
- The network achieved Dice coefficients of 75.7% (EX), 76.53% (SE), 50.06% (MA), and 45.89% (HE) on the IDRiD dataset.
- The multi-attention generative adversarial network demonstrated effective lesion segmentation.
Conclusions:
- The proposed generative adversarial network is effective for segmenting diabetic retinopathy in fundus images.
- This AI-driven approach shows promise for clinical applications in ophthalmology.
