Segmentation of Blood Vessels in 2D Retinal Images with Graphical Processing Units

Akca Z., Sen B., Akca E.

25th Signal Processing and Communications Applications Conference (SIU), Antalya, Türkiye, 15 - 18 Mayıs 2017 identifier identifier


Hypertensive Retinopathy (HR) cause serious damage to human body if not diagnosed early. Therefore, the early diagnosis of this anomaly is very important and complicated process. The most prominent aspect of HR is that the thickness of the blood vessels on the retina change. This change can be calculated with Arteriovenous ratio (AVR). Just before AVR calculation, the veins must be found and extracted on the retina image using some image processing techniques. In this study, the detection of vessels which have a very important place in the diagnosis of HR was performed on GPUs. The retina images taken from the DRIVE database were used as test data and the vessels were successfully detected in a much shorter time and without compromising sensitivity.