Keratoconus Diagnostic and Treatment Algorithms Based on Machine-Learning Methods

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Accurate keratoconus diagnosis is crucial for timely treatment and visual rehabilitation.
• Existing keratometry indices and classifications aid in quantifying disease severity.
• Advancements in computer processing and data analysis are enhancing diagnostic capabilities.

Purpose of the Study:

• To develop a machine-learning-based algorithm for precise keratoconus staging.
• To enable optimal patient management through accurate disease classification.
• To create a standardized keratoconus management algorithm based on predicted stages.

Main Methods:

• A multicentre retrospective study was conducted to build a patient database.
• Machine learning techniques, including principal component analysis and clustering, were employed.
• Keratotopographer readings were analyzed to classify disease severity.

Main Results:

• The algorithm accurately distinguishes between normal, preclinical, and stages 1-4 keratoconus.
• Performance was validated with an area under the curve (AUC) of 0.95 to 1.00.
• A web-based interface was developed for clinical application.

Conclusions:

• The machine learning algorithm offers precise keratoconus staging.
• This tool facilitates timely and appropriate management strategies for keratoconus patients.
• The developed software is suitable for clinical use, improving patient care.