Gradient based ant spread modification on ant colony optimization method for retinal blood vessel edge detection

Abstract

Retinal blood vessels are a very important object to diagnosing illness. Fast, automatic, adaptive, and accurate systems to segment the retinal vessels are particularly useful. The retinal blood vessels can be detected by edge detection. This research will compare the edge detection techniques with adaptive ant colony optimization (Adaptive ACO). Generally, the early ants on the conventional ant colony optimization (ACO) are randomly distributed. This condition can cause the ant distribution imbalances. Based on this problem, the ant distribution modification on ACO is proposed to optimize the ant placement based on the gradient. The gradient values are used to determine the ants' placement. The ants are not randomly distributed but placed in the highest gradient. The so-called Adaptive ACO is expected to be used for better and faster path discovery optimization. The average PSNR of Prewitt edges, Sobel edges, conventional ACO, and the proposed method are 11.605, 11.913, 9.783, and 15.874 respectively. The PSNR of the proposed method has the greatest value than others. It shows that the placement of ants based on gradient can improve edge detection accuracy. The application of the Adaptive ACO method has successfully optimized the result of retinal vessel edge detection.