Decision support system for age-related macular degeneration using discrete wavelet transform

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer-Aided Diagnosis

Background:

• Age-related macular degeneration (AMD) impacts central vision in individuals over 60.
• Dry AMD is the prevalent form, characterized by lesions like drusen.
• Manual detection of AMD lesions from retinal images is resource-intensive.

Purpose of the Study:

• To develop an automated system for detecting dry AMD.
• To utilize discrete wavelet transform (DWT) and feature ranking for AMD classification.
• To create an AMD Risk Index for simplified classification.

Main Methods:

• Feature extraction from DWT coefficients using statistical moments, energy, entropy, and Gini index.
• Application of five feature ranking strategies (t test, KLD, Chernoff Bound, Bhattacharyya Distance, ROC, Wilcoxon).
• Classification using SVM, decision tree, k-NN, Naive Bayes, and PNN.

Main Results:

• The system achieved 93.70% average accuracy, 91.11% sensitivity, and 96.30% specificity with KLD ranking and SVM.
• Optimal feature sets were identified using ranking strategies.
• An AMD Risk Index was formulated for classifying normal and dry AMD cases.

Conclusions:

• The proposed automated system effectively detects dry AMD.
• The system can assist clinicians in diagnosis and facilitate mass screening.
• This approach offers a more efficient alternative to manual lesion inspection.