Protocol for the diagnosis of keratoconus using convolutional neural networks

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Keratoconus is a prevalent corneal disease affecting 1 in 2000 individuals.
• Early diagnosis is critical for effective treatment outcomes.
• Current diagnostic tools provide parameters but lack comprehensive diagnostic capabilities.

Purpose of the Study:

• To develop a convolutional neural network (CNN) for the accurate diagnosis of keratoconus.
• To leverage a large dataset of patient information for training the diagnostic model.

Main Methods:

• Utilizing a comprehensive dataset of nearly 2000 patient records, including over 1000 longitudinal cases.
• Developing and training a convolutional neural network model for image-based diagnosis.
• Applying machine learning techniques to medical image analysis.

Main Results:

• The study focuses on developing a diagnostic tool rather than presenting specific results at this stage.
• The aim is to achieve a full image diagnosis capability for keratoconus.

Conclusions:

• A convolutional neural network holds promise for advancing keratoconus diagnosis.
• The developed model, trained on extensive patient data, aims to provide a crucial diagnostic tool for this common corneal disease.