Related Experiment Video
Updated: Oct 14, 2025
04:36
Author Spotlight: Understanding Retinal Vessel Resilience and Disease Progression
Published on: January 12, 2024
1.3K
A novel four-step feature selection technique for diabetic retinopathy grading
N Jagan Mohan1, R Murugan2, Tripti Goel1
1Bio-Medical Imaging Laboratory (BIOMIL), Department of Electronics and Communication Engineering, National Institute of Technology Silchar, Silchar, Assam, 788010, India.
Physical and Engineering Sciences in Medicine
|November 8, 2021
Summary
This study introduces an automated method for detecting diabetic retinopathy using deep learning and feature selection. The novel approach achieves high accuracy, aiding early diagnosis and preventing vision loss.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Diabetic retinopathy is a leading cause of vision loss and a microvascular complication of diabetes.
- Current diagnosis relies on time-consuming manual analysis by experts, necessitating automated solutions.
- Early detection is crucial to prevent irreversible vision impairment.
Purpose of the Study:
- To develop an automated system for accurate and efficient diabetic retinopathy detection.
- To propose a novel four-step feature selection technique to optimize diagnostic performance.
- To reduce human error and facilitate real-time screening for diabetic retinopathy.
Main Methods:
- An ensemble deep neural network (InceptionV3, ResNet101, Vgg19) was utilized for feature extraction from retinal fundus images.
- A four-step feature selection process involving minimum redundancy, maximum relevance, Chi-Square, ReliefF, and F-test was implemented.
- Support vector machines (SVM) were employed for the final classification of diabetic retinopathy.
Main Results:
- The proposed method achieved high diagnostic performance across multiple public datasets (Kaggle, MESSIDOR-2, IDRiD).
- An accuracy of 97.78%, sensitivity of 97.6%, and specificity of 99.3% were obtained using the top 300 selected features.
- The algorithm demonstrated superior performance compared to existing state-of-the-art methods.
Conclusions:
- The developed ensemble deep learning model with advanced feature selection offers a robust solution for automated diabetic retinopathy detection.
- This technique can significantly improve the efficiency and accuracy of screening, aiding in the prevention of vision loss.
- The findings support the integration of AI-powered tools in clinical practice for diabetic retinopathy management.
