A Clinically Explainable AI-Based Grading System for Age-Related Macular Degeneration Using Optical Coherence...

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Age-related macular degeneration (AMD) is a leading cause of vision loss.
• Accurate and timely diagnosis of AMD and its subtypes is crucial for effective treatment.
• Current diagnostic methods can be subjective and time-consuming.

Purpose of the Study:

• To develop an automated, explainable AI (xAI) system for diagnosing AMD from OCT images.
• To differentiate between normal retinas, various AMD grades, and non-AMD diseases.
• To extract and utilize clinically meaningful imaging markers for diagnosis.

Main Methods:

• Utilized optical coherence tomography (OCT) B-scan images.
• Developed an xAI system integrating DeepLabV3+ and a novel CNN model.
• Extracted key clinical imaging markers: subretinal/intraretinal fluid, choroidal hypertransmission, merged retinal layers, drusen, and retinal layer thickness.
• Employed a hierarchical decision tree for classification.

Main Results:

• The xAI system achieved 90.82% accuracy in a multi-way classification task on 1285 OCT images.
• Successfully differentiated normal retinas, early, intermediate, geographic atrophy (GA), and wet AMD, as well as non-AMD diseases.
• Demonstrated the system's capability to mimic physician perception in diagnosing AMD.

Conclusions:

• The proposed xAI system offers a promising automated approach for AMD diagnosis.
• The system can accurately classify various stages of AMD and distinguish from other retinal pathologies.
• Explainable AI holds potential for improving diagnostic efficiency and accuracy in ophthalmology.