A Hybrid Technique for Diabetic Retinopathy Detection Based on Ensemble-Optimized CNN and Texture Features

Area of Science:

• Ophthalmology and Medical Imaging
• Artificial Intelligence in Healthcare
• Computational Biology

Background:

• Diabetic retinopathy (DR) is a leading cause of irreversible vision loss in diabetic patients.
• Accurate staging of DR is critical for timely intervention and preventing severe visual impairment.
• Existing methods may lack the precision required for early and reliable DR detection.

Purpose of the Study:

• To develop and validate a novel hybrid method for the detection and classification of diabetic retinopathy stages.
• To differentiate between non-proliferative and proliferative stages of DR for effective patient management.
• To enhance the accuracy of DR classification using advanced machine learning techniques.

Main Methods:

• A hybrid approach combining image preprocessing with ensemble features derived from Local Binary Patterns (LBP) and deep learning.
• Development of a custom Convolutional Neural Network (CNN) model.
• Optimization of the ensemble feature vector using the Binary Dragonfly Algorithm (BDA) and Sine Cosine Algorithm (SCA).
• Classification using Support Vector Machine (SVM) on the optimized feature vector.

Main Results:

• The proposed hybrid methodology achieved a high classification accuracy of 98.85% on the Kaggle EyePACS dataset.
• The ensemble feature optimization using BDA and SCA significantly improved classification performance.
• Comparative analysis demonstrated the superiority of the proposed method over existing state-of-the-art approaches.

Conclusions:

• The developed hybrid approach offers a highly effective and accurate method for diabetic retinopathy stage classification.
• This AI-driven methodology holds significant potential for early detection and prevention of vision loss in diabetic patients.
• The integration of LBP, deep learning features, and metaheuristic optimization shows promise for advancing automated DR screening.