An ensemble classification of exudates in color fundus images using an evolutionary algorithm based optimal features...

Area of Science:

• Ophthalmology
• Medical Imaging
• Machine Learning

Background:

• Diabetic retinopathy (DR) detection relies on identifying exudates (EXs) in retinal images.
• Automated analysis of fundus images is crucial for early DR diagnosis.
• Feature selection is a key challenge for efficient and accurate EX classification.

Purpose of the Study:

• To propose a novel machine learning technique for early detection and classification of EXs in color fundus images.
• To optimize feature selection for reduced computational complexity and enhanced classification accuracy.
• To evaluate the performance of an ensemble classifier for EX detection.

Main Methods:

• An evolutionary algorithm was employed for optimal feature selection.
• Naïve Bayes, Support Vector Machine, and Artificial Neural Network classifiers were utilized.
• An ensemble-based classifier with majority voting was implemented for final classification.

Main Results:

• The proposed technique achieved 98% accuracy in detecting and classifying EXs.
• The evolutionary algorithm effectively reduced computational complexity.
• The ensemble classifier demonstrated superior performance in EX classification.

Conclusions:

• The developed machine learning technique offers a highly accurate method for automated EX detection in diabetic retinopathy.
• Optimal feature selection via evolutionary algorithms significantly improves classification efficiency.
• The ensemble approach provides a robust solution for classifying exudates in retinal images.