Related Experiment Video
Updated: Dec 25, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
Deep Transfer Learning Models for Medical Diabetic Retinopathy Detection
Nour Eldeen M Khalifa1, Mohamed Loey2, Mohamed Hamed N Taha1
1Information Technology Department, Faculty of Computers and Artificial Intelligence Cairo University, Giza, Egypt.
Summary
Early detection of diabetic retinopathy (DR) is crucial for preventing blindness. This study shows AlexNet achieved 97.9% accuracy in detecting DR using deep learning on the APTOS 2019 dataset.
Area of Science:
- Ophthalmology
- Computer Science
- Artificial Intelligence
Background:
- Diabetic retinopathy (DR) is a leading cause of blindness globally, affecting millions of diabetic patients.
- The incidence of diabetes is projected to rise significantly, increasing the prevalence of DR.
- Early detection of DR is critical to prevent vision loss.
Purpose of the Study:
- To investigate deep transfer learning models for early detection of diabetic retinopathy (DR).
- To evaluate the performance of various deep learning models on the APTOS 2019 dataset for DR detection.
- To identify the most effective model for accurate and efficient DR screening.
Main Methods:
- Employed deep transfer learning models including AlexNet, ResNet18, SqueezeNet, GoogleNet, VGG16, and VGG19.
- Trained and tested models on the Asia Pacific Tele-Ophthalmology Society (APTOS) 2019 dataset.
- Utilized data augmentation techniques to enhance model robustness and prevent overfitting.
Main Results:
- AlexNet achieved the highest testing accuracy of 97.9% among the evaluated models.
- Performance metrics including precision, recall, and F1 score validated the robustness of the models.
- AlexNet's minimal layer count reduced training time and computational complexity.
Conclusions:
- Deep transfer learning models, particularly AlexNet, demonstrate high accuracy in detecting diabetic retinopathy.
- The findings suggest AI-powered tools can significantly aid in early DR screening.
- This research provides a strong foundation for developing automated DR detection systems.
