Development of an application for providing corneal topography reports based on artificial intelligence

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Corneal topography is crucial for diagnosing various eye conditions.
• Current methods may require specialized equipment and complex analysis.
• Mobile technology offers potential for accessible eye diagnostics.

Purpose of the Study:

• To develop the TopEye iOS application for capturing and interpreting corneal topography maps using artificial intelligence.
• To create a user-friendly tool for mobile-based eye diagnostics.

Main Methods:

• Agile software development (Scrum) was employed.
• A diagnostic pattern bank of 1,172 corneal topography images was created, categorized into spherical, symmetrical, asymmetrical, and irregular (keratoconus) patterns.
• A neural network was trained on 960 images and tested on 212 images for diagnostic interpretation.

Main Results:

• The TopEye application demonstrated high diagnostic accuracy, correctly classifying 94.81% of 201 cases.
• The algorithm achieved 95.00% sensitivity and 98.68% specificity in keratoconus screening.
• The primary challenge identified was distinguishing between symmetrical and asymmetrical corneal patterns.

Conclusions:

• An efficient mobile application, TopEye, was developed for capturing and interpreting corneal topography images via smartphone.
• The application leverages artificial intelligence for semi-automatic diagnosis, improving accessibility.
• Further refinement may enhance accuracy in differentiating similar corneal patterns.