Local configuration pattern features for age-related macular degeneration characterization and classification

Area of Science:

• Ophthalmology and Medical Imaging

Background:

• Age-related Macular Degeneration (AMD) is a leading cause of irreversible vision loss in the elderly, affecting central vision due to macular cell degeneration.
• AMD presents in dry and wet forms, with dry AMD being more prevalent, and early detection is crucial to slow disease progression.

Purpose of the Study:

• To develop an automated system for diagnosing Age-related Macular Degeneration (AMD) using fundus images, aiming to reduce clinician screening time.
• To characterize normal and AMD classes by extracting and analyzing specific image features.

Main Methods:

• Extracted Linear Configuration Coefficients (CC) and Pattern Occurrence (PO) features from fundus images.
• Ranked features using p-value of t-test and employed various supervised classifiers including Decision Tree, k-NN, Naive Bayes, PNN, and SVM.
• Evaluated performance on private and public datasets (ARIA, STARE) using ten-fold cross-validation.

Main Results:

• The proposed approach achieved a highest average accuracy of 97.78%, sensitivity of 98.00%, and specificity of 97.50% on the STARE dataset.
• The system effectively classified normal and AMD classes using 22 significant features.

Conclusions:

• The developed automated AMD diagnosis system demonstrates high performance and reliability.
• This system can serve as a valuable aiding tool for clinicians in large-scale eye screening programs for early AMD detection.