Retinal Vessel Segmentation: An Efficient Graph Cut Approach with Retinex and Local Phase
April
Retinal Vessel Segmentation: An Efficient Graph Cut Approach with Retinex and Local Phase
Yitian Zhao 0 1 2 3
Yonghuai Liu 0 1 2 3
Xiangqian Wu 0 1 2 3
Simon P. Harding 0 1 2 3
Yalin Zheng 0 1 2 3
0 1 Department of Eye and Vision Science, University of Liverpool , Liverpool , United Kingdom , 2 Department of Computer Science, Aberystwyth University , Aberystwyth , United Kingdom , 3 School of Computer Science and Technology, Harbin Institute of Technology , Harbin, China, 4 St Paul's Eye Unit , Royal Liverpool University Hospital , Liverpool , United Kingdom
1 Funding: This work was supported by the Wellcome Trust (grant number 092668/Z/10/Z). The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript
2 Academic Editor: Stefan Strack, University of Iowa, UNITED STATES
3 Current address: School of Mechatronical Engineering, Beijing Institute of Technology , Beijing , China
Our application concerns the automated detection of vessels in retinal images to improve understanding of the disease mechanism, diagnosis and treatment of retinal and a number of systemic diseases. We propose a new framework for segmenting retinal vasculatures with much improved accuracy and efficiency. The proposed framework consists of three technical components: Retinex-based image inhomogeneity correction, local phase-based vessel enhancement and graph cut-based active contour segmentation. These procedures are applied in the following order. Underpinned by the Retinex theory, the inhomogeneity correction step aims to address challenges presented by the image intensity inhomogeneities, and the relatively low contrast of thin vessels compared to the background. The local phase enhancement technique is employed to enhance vessels for its superiority in preserving the vessel edges. The graph cut-based active contour method is used for its efficiency and effectiveness in segmenting the vessels from the enhanced images using the local phase filter. We have demonstrated its performance by applying it to four public retinal image datasets (3 datasets of color fundus photography and 1 of fluorescein angiography). Statistical analysis demonstrates that each component of the framework can provide the level of performance expected. The proposed framework is compared with widely used unsupervised and supervised methods, showing that the overall framework outperforms its competitors. For example, the achieved sensitivity (0:744), specificity (0:978) and accuracy (0:953) for the DRIVE dataset are very close to those of the manual annotations obtained by the second observer.
-
Competing Interests: The authors have declared
that no competing interests exist.
vasculature can result from several diseases [1]. Vascular abnormalities can be seen in various
retinal diseases. Study of the retinal circulation is of great importance in the management of
retinal diseases, but also provides unique opportunity to study the microvascular damage to
the brain in cerebral malaria [2]. Structural changes in the retinal vasculature may also indicate
hypertension, stroke, heart disease and nephropathy [3]. The retina is visible to examination
and accessible to high-resolution, non-invasive imaging. This provided a unique window that
allows direct visualization and analysis of the inner retinal vascular circulation for studying
various related conditions. Automated analysis of the retinal vasculature becomes an active
research area in the field of medical imaging for its diagnostic and prognostic significance.
Our application concerns the automated detection of retinal blood vessels in diagnostic
retinal images such as color fundus images and fluorescein angiography images. The automated
detection of blood vessels is a prerequisite in the development of automated system for the
analysis of vessels. Recent years have witnessed the rapid development of methods for retinal
vessel segmentation, as evidenced by extensive reviews [4, 5]. For the purpose of this paper this
list is intended only to provide readers with some insight into this problem domain, and is by
no means exhaustive. Most existing methods are automated techniques without interaction
from the user during the segmentation. However, we noted that interactive segmentation
techniques, such as Live Vessel [6], were proposed for improving the segmentation performance.
Broadly speaking, all the established automated segmentation techniques may be categorized
as either supervised segmentation [713] or unsupervised segmentation [1423] with respect
to the overall system design and architecture.
Supervised segmentation requires hand-labeled gold standard images for training, and each
pixel is represented by a feature vector which is obtained from local or global information of
the image. The prerequisite for this approach is that a set of features having the necessary
discriminative ability have to be extracted for training and classificati (...truncated)