Applications of machine learning in glaucoma diagnosis based on tabular data: a systematic review

Area of Science:

• Ophthalmology
• Medical Informatics
• Artificial Intelligence

Background:

• Glaucoma is a primary cause of irreversible blindness globally.
• Current diagnostic methods for glaucoma lack objectivity and consistency.
• There is a critical need for advanced, reliable diagnostic tools in ophthalmology.

Purpose of the Study:

• To systematically review and evaluate the efficacy of machine learning (ML) techniques for glaucoma diagnosis.
• To identify the most effective ML models and data types for improving glaucoma detection accuracy.
• To assess the challenges and future directions for ML application in clinical glaucoma diagnosis.

Main Methods:

• A systematic literature review was performed across five major databases.
• 35 relevant studies were selected based on predefined inclusion and exclusion criteria.
• Analysis focused on ML model performance metrics (accuracy, AUC) using structured data (OCT, VF, demographics).

Main Results:

• Structured data, including optical coherence tomography (OCT), visual field (VF) tests, and demographic information, significantly boosted diagnostic accuracy.
• Machine learning models, particularly support vector machine (SVM), deep learning (DL), random forest, and ensemble methods, achieved accuracies from 76% to 98.3%.
• Area Under the Curve (AUC) values for these models ranged from 52.5% to 99%, indicating varying degrees of diagnostic capability.

Conclusions:

• Machine learning holds significant potential to enhance the accuracy and reliability of glaucoma diagnosis.
• Challenges such as data imbalance and limited sample sizes currently hinder the generalizability of ML models.
• Further research focusing on data quality improvement and robust model validation is essential for widespread clinical adoption of ML in glaucoma detection.