Classification Algorithms Used in Predicting Glaucoma Progression.

Area of Science:

• Ophthalmology
• Medical Informatics
• Artificial Intelligence

Background:

• Open-angle glaucoma (POAG) is a leading cause of irreversible blindness worldwide.
• Accurate prediction of POAG progression is crucial for timely intervention and patient management.
• Existing methods for predicting glaucoma progression have limitations in accuracy and interpretability.

Purpose of the Study:

• To evaluate the efficacy of various machine learning algorithms in predicting the progression of open-angle glaucoma.
• To compare the performance of different classification algorithms using clinical datasets.
• To identify algorithms that provide high accuracy and interpretable insights into glaucoma progression.

Main Methods:

• Utilized datasets from UK and Romanian hospitals containing clinical observations and objective measurements.
• Employed machine learning algorithms including Multilayer Perceptron, Random Forest, C4.5, k-Nearest Neighbors, Support Vector Machine, and Non-Nested Generalized Exemplars.
• Evaluated glaucoma progression based on Visual Field Index (VFI), Mean Deviation (MD), Pattern Standard Deviation (PSD), and Retinal Nerve Fiber Layer (RNFL) parameters.

Main Results:

• Multilayer Perceptron and Random Forest algorithms achieved prediction accuracy exceeding 90%.
• The Non-Nested Generalized Exemplars (NNGE) algorithm effectively determined the influence of input variables on outcomes.
• Decision tree algorithms provided valuable insights into classification logic, aiding in understanding decision-making processes.

Conclusions:

• Machine learning, particularly Multilayer Perceptron and Random Forest, offers a highly accurate approach for predicting POAG progression.
• Interpretable models like decision trees and NNGE enhance clinical understanding and trust in AI-driven predictions.
• These findings support the integration of machine learning tools for improved glaucoma diagnosis and management.