[Artificial intelligence for eye care]

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Technological advancements in ophthalmic imaging and AI offer new diagnostic capabilities in eye care.
• AI is already utilized in ophthalmic diabetes care, with systems demonstrating high sensitivity and specificity in detecting diabetic retinopathy.
• AI systems for screening, monitoring, and treating age-related macular degeneration and glaucoma are under development.

Purpose of the Study:

• To explore the applications and challenges of AI in ophthalmic diagnostics.
• To discuss the performance dependency of AI algorithms on training data and reference standards.
• To address the ongoing issue of interpretability in deep learning algorithms for medical imaging.

Main Methods:

• Review of current AI applications in ophthalmic diagnostics.
• Analysis of factors influencing AI algorithm performance, including data and gold standards.
• Discussion of interpretability methods for AI in medical imaging.

Main Results:

• AI systems achieve high sensitivity and specificity in detecting diabetic retinopathy.
• AI shows promise for age-related macular degeneration and glaucoma, with ongoing development.
• Algorithm performance is critically dependent on the quality and establishment of training data and reference standards.

Conclusions:

• AI is a powerful tool for ophthalmic diagnostics, particularly in diabetic retinopathy detection.
• Further research is needed to improve AI interpretability and address data dependency issues.
• Visualizing critical image areas can enhance user understanding of AI-driven diagnostic decisions.