Age-related macular degeneration diagnosis in optical coherence tomography images with gray level co-occurrence matrix features, genetic algorithms, and random forest classifier.

Abstract

This paper proposes new computational strategies to improve optical coherence tomography image quality for age-related macular degeneration identification. Integrated into an age-related macular degeneration detection system, these algorithms automate and improve the identification of abnormalities in optical coherence tomography images, aiding in the classification of normal and abnormal macular tissues. The research presents an innovative approach to detecting age-related macular degeneration related anomalies, combining texture analysis, statistical evaluation, and genetic algorithms for feature selection. Genetic algorithm optimization finds the best predictive characteristics by using in-depth texture analysis with the gray level co-occurrence matrix and comprehensive statistical research. Gray level co-occurrence matrix features are analyzed at four angles (0°, 45°, 90°, and 135°), with the random forest classifier trained using optimized features. The random forest classifier plays a vital role in both the training and testing phases, achieving no-table results: an error rate of 0% for selected features, 1.9% for all features, and 7.5% for no features, and an overall system classification accuracy of 100% for training data for all, while maintaining 92.458% 98.113%, 100% for testing data of no features, all features and selected features.