Artificial intelligence support improves diagnosis accuracy in anterior segment eye diseases

Area of Science:

• Ophthalmology
• Medical Artificial Intelligence
• Diagnostic Imaging

Background:

• Ophthalmologists face challenges in accurately diagnosing a range of corneal diseases and cataracts.
• Deep learning models offer potential solutions to augment diagnostic capabilities in clinical practice.

Purpose of the Study:

• To evaluate the impact of CorneAI, a deep learning model, on the diagnostic accuracy of ophthalmologists for various eye conditions.
• To compare diagnostic performance with and without AI assistance across different imaging modalities.

Main Methods:

• 40 ophthalmologists (20 specialists, 20 residents) classified 100 eye images (50 smartphone, 50 slit-lamp) into nine categories.
• Participants initially diagnosed cases without CorneAI, then re-evaluated the same cases with CorneAI assistance 2-4 weeks later.

Main Results:

• Overall ophthalmologist diagnostic accuracy increased significantly from 79.2% to 88.8% with CorneAI (P < 0.001).
• Accuracy improved for both specialists (82.8% to 90.0%) and residents (75.6% to 86.2%).
• CorneAI enhanced accuracy for both smartphone (78.7% to 85.5%) and slit-lamp images (81.2% to 90.6%).

Conclusions:

• CorneAI effectively enhances ophthalmologists' diagnostic accuracy for cataracts and corneal diseases.
• The AI model improves diagnostic performance regardless of image source, including smartphone-captured images.
• AI-assisted diagnosis shows significant benefits for both experienced specialists and residents.