Performance comparison of artificial bee colony algorithm based approaches for retinal vessel segmentation

Structural changes in the retinal blood vessels provide important information about retinal diseases. Therefore, computer-aided segmentation of retinal blood vessels has become an active area of research in last decades. Due to the close contrast between the retinal blood vessels and the retinal background, robust methods should be developed to detect retinal blood vessels with high accuracy. In this work, artificial bee colony (ABC) algorithm which provides effective solutions to engineering problems has been applied to the retinal vessel segmentation. Clustering based ABC (basic ABC), quick-ABC (Q-ABC) and modified ABC (MR-ABC) algorithms have been analyzed for accurate segmentation of retinal blood vessels and their performances were compared. The simulations have been realized on the normal and abnormal retinal images taken from the DRIVE database. Simulation results and statistical analyses represent that ABC based approaches are stable and able to reach to optimal clustering performance with higher convergence rates. As a result it can be concluded that ABC based approaches can successfully be used for accurate segmentation of retinal blood vessels.