A Novel Machine Learning-Based Classification Framework for Age-Related Macular Degeneration (AMD) Diagnosis from Fundus Images

• Main Authors: El-Khalek, Aya A. Abd, Balaha, Hossam Magdy, Mahmoud, Ali, Alghamdi, Norah Saleh, Ghazal, Mohammed, Khalil, Abeer T., Abo-Elsoud, Mohy Eldin A., El-Baz, Ayman
• Format: Conference Proceeding
• Language: English
• Subjects: Accuracy; Age-Related Macular Degeneration (AMD); Computer Aided Diagnosis (CAD); Feature extraction; Machine Learning (ML); Macular degeneration; Optical imaging; Retina; Sensitivity; Solid modeling; Weighte Majority Voting (WMV)
• ISSN: 1945-8452
• DOI: 10.1109/ISBI56570.2024.10635727

The increasing prevalence of eye disorders among the elderly has underscored the importance of early detection through routine eye examinations. Age-related Macular Degeneration (AMD), a common ailment affecting those over 45, is a leading cause of vision impairment in older individuals. This study introduces a comprehensive Computer-Aided Diagnosis (CAD) system that leverages ML techniques to categorize fundus images into geographic atrophy (GA), intermediate AMD, normal, and wet AMD categories. We developed an innovative system that extracts both local and global appearance markers from fundus images. These markers are derived from the entire retina and iso-regions aligned with the optical disc, utilizing the fast marching level sets model. The features include markers from the Gray-Level Co-occurrence Matrix (GLCM) and Gray-Level Run-Length Matrix (GLRLM) models. The integration of weighted majority voting with the best classifiers enhances performance, resulting in an impressive overall accuracy of 96.85%, sensitivity of 93.72%, specificity of 97.89%, precision of 93.86%, F1 score of 93.72%, ROC of 95.85%, BAC of 95.81%, and WSM of 95.38%.