Glaucoma risk assessment based on clinical data and automated nerve fiber layer defects detection

Area of Science:

• Ophthalmology
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Glaucoma is a leading cause of irreversible vision loss, particularly in Japan.
• Automated detection of nerve fiber layer defects (NFLDs) exists but its utility for glaucoma risk assessment is unevaluated.
• Accurate and early glaucoma diagnosis is critical for effective management and prevention of blindness.

Purpose of the Study:

• To develop and evaluate a computerized glaucoma risk assessment system.
• To compare the performance of this risk assessment with automated NFLD detection.
• To determine the utility of machine learning models integrating clinical data for glaucoma risk stratification.

Main Methods:

• Utilized machine learning techniques including artificial neural networks, RBF networks, k-nearest neighbors, and support vector machines.
• Developed a glaucoma risk assessment model using ten clinical parameters, with and without NFLD detection results.
• Included systemic data, ophthalmologic data, and retinal images in the clinical dataset.

Main Results:

• The developed computerized glaucoma risk assessment demonstrated superior performance compared to NFLD detection alone.
• Machine learning models effectively integrated clinical information for predicting glaucoma risk.
• The system showed potential for accurate glaucoma risk determination.

Conclusions:

• Computerized glaucoma risk assessment using machine learning is a promising tool for clinical application.
• Integrating diverse clinical data enhances the accuracy of glaucoma risk prediction.
• This approach may significantly aid in the early determination of glaucoma risk, potentially reducing vision loss.