Identifying Vision Disorders Using Pupil Color Analysis

Abstract

Amblyopia is a neurological vision disorder that studies show affects two to five percent of the population. Current methods of treatment produce the best visual outcome if the condition is identified early in the patient's life. Several early screening procedures are aimed at finding the condition while the patient is a child, including an automated vision screening system developed by Cibis, Wang, and Van Eenwyk. The system uses artificial intelligence software algorithms to achieve a 77% accuracy in identifying patients who are at risk for developing the amblyopic condition and should be referred to a specialist. This thesis aims to improve upon the existing automated vision screening system and increase the sensitivity, specificity, and accuracy measurements. It explores the application of decision tree learning algorithms and artificial neural networks on a previously unused set of features. The features are extracted from images of patient eyes and focus on the color information contained. The efficacy of pixel color data is also investigated with respect to the measurement of the rate of change of the color in the iris and pupil. Processing the data and testing the machine learning applications using a 10-fold stratified cross validation procedure reveals that the best results show an overall accuracy of 68% in identifying patients who are at risk of developing the amblyopic condition. These results do not outperform the previous research; however, the process has allowed an in-depth investigation into the potential of the iris and pupil color slope features.