MACHINE LEARNING MODELS FOR PREDICTION OF GLAUCOMA STATUS USING SIGNIFICANT DEMOGRAPHIC, CLINICAL, AND LIFESTYLE RISK FACTORS

Abstract

Glaucoma is an eye condition that damages the optic nerve, leading to irreversible vision loss. Despite advances in diagnostic techniques, it remains a significant public health concern, particularly in resource-constrained settings. Previous studies on Machine Learning (ML) used Demographic Risk Factors (DRFs), Clinical Risk Factors (CRFs), and Fundus Images (FI) to predict Glaucoma, but did not emphasize the use of lifestyle risk factors (LRFs) alongside DRFs and CRFs. A dataset of sample size 200 patients was collected from the National Eye Centre (NEC), Kaduna, from 25th November 2024 to 19th December 2024 using personal interview. The results of the study revealed that all fourteen risk factors of Glaucoma were significant. SVM, DT, K-NN, NB, and MLP ML models predicted 97.1%, 99.0%, 96.1%, 96.1%, 98.0% and 92.6%, 100%, 92.6%, 96.3%, 96.3% Glaucoma patients in the training and test sets.  94.8%, 98.3%, 94.8%, 93.1%, 96.6% and 84.6%, 92.3%, 76.9%, 84.6%, 92.3% Non-Glaucoma patients in the training and test sets. The models have perfect performance and better 5-folds and 8-folds cross-validation. The study concludes that the use of significant LRFs alongside with DRFs and CRFs could help to predict Glaucoma and Non-Glaucoma patients effectively.