Automated Glaucoma Detection And Classification In Digital Ocular Fundus Images

Abstract

— The upward trend in the number of people suffering from Diabetes has led to a surge in the
number of working adults suffering from Glaucoma in several developed countries. To date, manual
screening of Glaucoma is still preferred as to automated screening in most of those countries. Coupled with
the lack of ophthalmologists for the increasing number of patients, manual screening is very tedious and
time- consuming. Hence, this paper proposes an automated glaucoma detection system in the view of
improving the accuracy of the previously developed systems. The system was developed using Adaptive
Thresholding, Mathematical Morphology and Grey Level Co-Occurrence Matrix for exudate detection and a
cascading neural network for exudate classification between Haemorrhages and Microaneurysms. The
neural network was trained using 100 ocular fundus images while the proposed system was tested using 25
images, obtained from DRIVE database. The system implemented yielded accuracy of 92% or 23 out of 25
images.