Automated measurement of retinal vessel diameters on digital fundus photographs

摘要

Recent studies of retina discovered that myopic refraction has a correlation with smaller retinal vessel diameters and variation in retinal arteriolar caliber is an important sign of systemic hypertension. Also, lower retinal arteriole to venule ratio (AVR) can predict the risk of hypertension.To utilize these valuable information in prognosing related diseases more precisely and quickly, this paper suggests an automated measurement algorithm of retinal vessel diameters. In this method, there is no requirement for human-intervention such as classification of vessels by hand or initial information of each vessel's position. This method is focused on a region around the optic disc for accurate measurements of retinal vessel's diameters. Region of donut shape, which is centered at optic disc, has clearer contrast than other region and make it possible to classify vessels into two types, arteriole and venule reliably.From the green channel image data, vessel's diameters can be automatically measured by using intensity profile graphs and significant features of retina, and it will help to obtain more accurate AVR, variance and mean values of width in the region of interest.