Hardware Implementation of Detection of Glaucoma from Color Fundus Images

Glaucoma is one of the most commonly observed retinal disease in India that leads to blindness if not diagnosed and treated at early stage. There are no specific symptoms for this disease, it is observed by loss of side vision. Glaucoma is a slow progressive degeneration of retinal ganglion cells (RGC) and their axons, resulting in a distinct appearance to the optic nerve head (ONH), often called cupping. Due to cupping the cup area increases and causes loss of side vision. Usually specially trained clinicians manually grade the fundus images in a time-consuming manner. The authors have developed a novel algorithm for automatic detection of AMD using image processing techniques and implementation on hardware. Glaucoma is assessed by estimating the cup to disc ratio (CDR). The proposed algorithm employed morphological and histogram techniques to identify cup and disc on fundus images and to measure the radius and area of cup and disc of the retina. Intel’s Open CV Library is used to implement the algorithm. The hardware is a Texas Instruments (TI) DM3730 based system on chip (SOC) low cost, low power single board computer system. The software developed is embedded on the hardware to test retinal images for the detection of AMD. The algorithm has been tested with 55 normal and 45 glaucomatous images and has yielded an accuracy of 94%.