A HW/SW embedded system for accelerating diagnosis of glaucoma from eye fundus images

Abstract

Glaucoma is an irreversible eye disease in which the optic nerve is progressively damaged leading to blindness. However, it is manageable if diagnosed early. The most common screening exam for glaucoma diagnosis is the eye fundus evaluation in which the ophthalmologist examines the optic nerve and estimates the Vertical Cup to Disc Ratio (VCDR). Currently, VCDR evaluation is performed by ophthalmologists based on their visual perception and experience. This paper explores different embedded architectures based on low power processors and describes a HW/SW embedded system that automatically calculates VCDR from eye fundus images using image processing techniques. Optic disc diameter is calculated by a HW accelerator, while optic cup diameter is calculated by SW. This resulted in an embedded system that reduces at least 30% of the execution time of a SW-only implementation and that is significantly faster than other related works based on desktop computers. The proposed system was tested on 70 eye fundus images and achieved a 97.72% accuracy rate.