AUTOMATIC EXTRACTION OF RETINAL BLOOD VESSELS: A SOFTWARE IMPLEMENTATION

Abstract

In the field of sports injuries, the injuries are often classified as traumatic or overuse injury. Also, high-velocity flying objects can do irreparable damage to the human eye. In traumatic head injuries, small blood vessels of athlete's eye are often damaged, also these vessels may produce bleeding within the front chamber of the eye, between the iris and cornea (e.g., Hyphema). In the scope of image processing, segmentation of the optic disc, blood vessels and macula in digital fundus images is important for the research area of medical image analysis. It is used to efficiently implement the diagnostic evaluation and taken as a basis of the cure to illness of patient. Automatic segmentation of retinal blood vessels is a preliminary step in the development of a computer-assisted diagnostic system for ophthalmic problems. The Sobel and Prewitt edge detectors are based on the discrete differentiation operators. The Kirsch compass kernel is a non-linear edge detector. These edge detectors usually find the maximum edge strength in predetermined directions. SCILAB is an open source, cross-platform numerical computational package. It uses a high-level, numerically oriented programming language that it is similar to MATLAB numerical computing environment. In this paper, the Sobel, Prewitt edge detectors and the edge detection based on the Kirsch templates (i.e., kernel) are used to detect and extract the retinal blood vessels by using developed SCILAB implementation. Considering the computer science in sport research area, the proposed approach given in this study aims to show an open source and efficient implementation of automatic blood vessel extraction, and comparison of the performance of three different edge detectors. In the experiments, the DRIVE retinal image database is used to test the developed approach. Experimental results show that our implementation based on Kirsch templates is more preferable than standard implementations of Sobel and Prewitt edge detectors used in SCILAB's image processing design (IPD) and image and video processing (SIVP) toolboxes.