Machine Learning-Based Analysis of Optical Coherence Tomography Angiography Images for Age-Related Macular...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Age-related macular degeneration (AMD) is a primary cause of vision loss in older adults.
• Optical coherence tomography angiography (OCTA) provides detailed retinal vascular imaging but presents diagnostic challenges.
• High data volume and subtle abnormalities in OCTA images complicate clinical assessment.

Purpose of the Study:

• To develop automated algorithms for detecting and quantifying AMD in OCTA images.
• To reduce ophthalmologists' workload and improve diagnostic accuracy for AMD.
• To create efficient tools for analyzing complex OCTA data.

Main Methods:

• Development of two texture-based algorithms for OCTA image classification without segmentation.
• Utilizing local texture descriptors (LBP2riu, LBP, BRIEF) with machine learning classifiers (SVM, KNN).
• Application of Principal Component Analysis (PCA) for feature reduction in the second algorithm.

Main Results:

• The first algorithm achieved perfect discrimination (AUC 1.00±0.00) between healthy eyes and wet AMD.
• The second algorithm demonstrated high performance in differentiating dry AMD from wet AMD (AUC 0.85±0.02).
• Evaluation performed on OCTA datasets from multiple eye hospitals.

Conclusions:

• The developed algorithms show significant potential for rapid and accurate AMD diagnosis using OCTA.
• Automated analysis can reduce manual evaluation variability and workload for clinicians.
• These tools may enhance clinical decision-making and patient management for AMD.