Related Experiment Video
Updated: Aug 9, 2025
05:54
Quantification of Diabetes-induced Adherent Leukocytes in Retinal Vasculature
Published on: January 24, 2025
362
EfficientNetV2 Based Ensemble Model for Quality Estimation of Diabetic Retinopathy Images from DeepDRiD
Sudhakar Tummala1, Venkata Sainath Gupta Thadikemalla2, Seifedine Kadry3,4,5
1Department of Electronics and Communication Engineering, School of Engineering and Sciences, SRM University-AP, Amaravati 522240, Andhra Pradesh, India.
Diagnostics (Basel, Switzerland)
|February 25, 2023
Summary
Automated quality estimation (QE) of retinal fundus images using EfficientNetV2 deep neural networks improves diabetic retinopathy (DR) screening accuracy. This method offers a reliable tool for ophthalmologists, reducing diagnostic errors in DR detection.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Diabetic retinopathy (DR) is a leading cause of blindness, diagnosed via retinal fundus images.
- Manual DR screening is time-consuming and prone to errors.
- High-quality fundus images are crucial for accurate DR diagnosis.
Purpose of the Study:
- To develop an automated method for estimating the quality of digital fundus images.
- To enhance the efficiency and accuracy of diabetic retinopathy screening.
Main Methods:
- An ensemble of EfficientNetV2 deep neural network models was employed for automated quality estimation (QE).
- The method was validated on the Deep Diabetic Retinopathy Image Dataset (DeepDRiD).
Main Results:
- The proposed automated QE method achieved a test accuracy of 75% on the DeepDRiD dataset.
- This performance surpasses existing methods for fundus image quality assessment.
Conclusions:
- The automated ensemble method is a promising tool for fundus image QE.
- This technology can assist ophthalmologists in efficient and accurate diabetic retinopathy screening.
