Diagnosis of Glaucoma Through the Analysis of Saccadic Eye Movements Employing Machine Learning Methods

This study addresses the significant public health concern of glaucoma, a collection of eye conditions resulting in optic nerve damage and progressive vision loss. To conduct our research, we employed a dataset obtained from collaborative efforts between the Amrita School of Engineering in Bangalore and Narayana Nethralaya Hospital during visual field examinations. The dataset, named EXGP, comprises Humphrey Visual Field tests conducted on a sample of 111 patients. Our research investigates the impact of variations in saccadic eye movements on glaucoma performance. This hypothesis is tested through an examination of diverse metrics associated with saccadic eye movements during visual search tasks. The summarizing parameter, Visual Field Index (VFI), is employed. However, for its widespread acceptance, further characterization and comparison with conventional indices used in glaucoma diagnosis, specifically Mean Deviation (MD), are required. Our objective is to define the applicability of VFI in advanced glaucoma cases by contrasting it with MD and the criteria for establishing blindness. Through statistical analysis of confidential eye data, we generated data samples from the provided dataset. Subsequently, a range of machine learning algorithms, including Support Vector Machine (SVM), K nearest neighbors (Knn), and Logistic Regression (LR), were utilized to predict the presence of glaucoma or normality in individuals. Classification accuracy for the various machine learning models was assessed using appropriate evaluation metrics. Among the three models, Logistic Regression demonstrated the highest accuracy of 0.94 in performing the classification task.