Development of glaucoma predictive model and risk factors assessment based on supervised models.

Area of Science:

• Ophthalmology
• Medical Informatics
• Machine Learning

Background:

• Glaucoma is a leading cause of irreversible blindness globally.
• Early detection and risk factor identification are crucial for managing glaucoma.
• Existing diagnostic methods can be invasive or require specialized equipment.

Purpose of the Study:

• To develop and propose machine learning models for glaucoma prediction.
• To identify significant risk factors associated with glaucoma development.
• To evaluate the performance of various machine learning algorithms in glaucoma detection.

Main Methods:

• Utilized the Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology.
• Employed data sampling (over-sampling and under-sampling), imputation, and normalization for preprocessing.
• Developed and compared Decision Trees, K-Nearest Neighbors, Support Vector Machines, Random Forests, Extra Trees, Bagging Ensembles, and a novel stacking ensemble model.

Main Results:

• Random Forests and Decision Trees, trained with under-sampling, demonstrated superior performance in predicting glaucoma.
• Random Forests achieved an accuracy of 88.87% and an Area Under the Curve (AUC) of 94.53%.
• Key predictors identified include visual field parameters, vertical cup-to-disk ratio, axial length, and systolic blood pressure.

Conclusions:

• Machine learning models can effectively predict glaucoma in individuals aged 40-64.
• The study identified crucial features for distinguishing glaucoma patients from non-glaucoma individuals.
• The developed models offer a potential tool for early glaucoma screening and risk assessment.