Related Experiment Video
Updated: Aug 9, 2025
09:17
Using Retinal Imaging to Study Dementia
Published on: November 6, 2017
21.7K
Triple-DRNet: A triple-cascade convolution neural network for diabetic retinopathy grading using fundus images.
Muwei Jian1, Hongyu Chen2, Chen Tao2
1School of Information Science and Technology, Linyi University, Linyi, China; School of Computer Science and Technology, Shandong University of Finance and Economics, Jinan, China.
Computers in Biology and Medicine
|February 22, 2023
Summary
A new deep learning model, Triple-DRNet, accurately grades Diabetic Retinopathy (DR) by classifying lesions. This automated approach improves early detection and treatment of DR, a leading cause of blindness.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Diabetic Retinopathy (DR) is a major cause of blindness globally, necessitating accurate and timely grading for effective treatment.
- Traditional Convolutional Neural Networks (CNNs) struggle to differentiate specific lesions, impacting the accuracy of DR classification.
Purpose of the Study:
- To develop an automated system for precise Diabetic Retinopathy grading.
- To enhance the classification of DR by effectively distinguishing various lesion types.
Main Methods:
- A novel triple-cascade network model, Triple-DRNet, was proposed for DR grading.
- The model employs a staged approach: DR vs. No DR, Proliferative DR (PDR) vs. Non-Proliferative DR (NPDR), and severity grading within NPDR (mild, moderate, severe).
Main Results:
- Triple-DRNet achieved an accuracy (ACC) of 92.08% on the APTOS 2019 Blindness Detection dataset.
- The model reached a Quadratic Weighted Kappa (QWK) metric of 93.62%, demonstrating superior performance.
Conclusions:
- The Triple-DRNet model significantly improves DR grading performance compared to existing methods.
- This cascaded network approach offers a more effective solution for automated DR classification and early intervention.
