Article Preview
Top1. Introduction
A new study finds that certain morphological changes of the retinal blood vessels in retinal images are important indicators for diseases like diabetes, hypertension and glaucoma. However, the length, width, diameter of the blood vessels may change as indications of various ophthalmologic diseases (Mendonca, 2006). For example, vessel occlusion that makes veins longer, abnormal narrowing of retinal blood vessels may indicate the earlier stages of glaucoma or DR that creates new blood vessels (neovascularization) which are tiny, thin, fragile and abnormal in nature and cause frequent minor bleeding that may lead to permanent vision loss (Gonzalez, 2014). Automatic extraction of retinal blood vessels is very important for the diagnosis and the treatment of different diseases such as hypertension (Leung, 2003), obesity, glaucoma (Wang, 2006) and diabetic retinopathy (Morello, 2007). Diabetic retinopathy is an increasingly growing public health problem and the leading cause of blindness in the world (Elbalaoui, 2014). In all cases, correct detection of retinal blood vessel is crucial. In most of the retinal screening programs, blood vessels are extracted manually. Figure 1 shows the eyeball structure and retinal image captured through a retinal imaging device.
Figure 1. (a) Anatomy of the eye, (b) Fundus image
Many methods can be found in the literature for segmentation of blood vessels, which can be divided into the following major categories: supervised and unsupervised methods.
Supervised methods require a feature vector for each pixel and manually labeled images in order to discriminate between vessel and non-vessel pixels. Soares et al. (Soares, 2006) used a Bayesian classifier with class-conditional probability density functions (likelihoods) described as Gaussian mixtures, yielding a classification to model complex decision surfaces. The probability distributions are estimated based on training set of labeled pixels obtained from manual segmentations. Xu et al. (Xu, 2011) proposed a method to segment retinal blood vessels to overcome the variations in contrast of large and thin vessels. This method uses adaptive local thresholding to produce a binary image then extract large connected components as large vessels. The residual fragments in the binary image including some thin vessel segments (or pixels), are classified by Support Vector Machine (SVM). The tracking growth is applied to the thin vessel segments to form the whole vascular network. Fraz et al. (Fraz, 2012) employed a classification system based on an ensemble of bagged and boosted decision trees and utilizes a feature vector based on the orientation analysis of gradient vector field, morphological transformation, line strength measures, and Gabor filter responses. Marin et al. (Marin, 2011) used a neural network (NN) scheme for pixel classification and computes a 7-D vector composed of gray-level and moment invariants-based features for pixel representation.
Unsupervised methods in the literature comprise the matched filter responses, grouping of edge pixels, adaptive thresholding, vessel tracking and morphology based techniques. Mendonça et al. (Mendonça, 2006) proposed an algorithm starts with the extraction of vessel centerlines, which are used as guidelines for the subsequent vessel filling phase. For this purpose, the outputs of four directional differential operators are processed in order to select connected sets of candidate points to be further classified as centerline pixels using vessel derived features.