Published on

Published in: Technology, Health & Medicine
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide


  1. 1. K. Sureka et al Int. Journal of Engineering Research and Applications www.ijera.com ISSN : 2248-9622, Vol. 4, Issue 3( Version 4), March 2014, pp.01-05 www.ijera.com 1 | P a g e A Vessel Tracking System for the Robust Extraction of Vascular Network Connectivity in Retinal Fundus Images K. Sureka1 , M.E., R. Vignesh2 , M.E 1 Student/Dept. of Applied Electronics 2 Assistant Professor/ECE Jayam College of Engineering and Technology, Dharmapuri DT, India. Abstract Blood vessel morphology is an important indicator for diseases like cardiovascular, hypertension and diabetic retinopathy. The wrong identification of vessels may result in a large variation of these measurements, leading to a wrong clinical diagnosis. The problem of identifying true vessels as a post- processing step to vascular structure segmentation. The segmented vascular structure as a vessel segment graph and formulate the problem of identifying vessels as one of finding the optimal forest in the graph given a set of constraints. Automatic method to detect blood vessel crossovers and bifurcations simultaneously. Detection is performed by interactive segmentation using graph cut algorithm to find all potential abnormalities. Vessel segmentation is the most important step for accurate and efficient vascular feature analysis to achieve high pixel precision of the true vessels for clean segmented retinal images. Index Terms: retinal image analysis, vascular structure, vessel identification, segmentation I. INTRODUCTION A Retinal image provides a snapshot of what is happening inside the human body. In practice, the state of the retinal vessels has been shown to reflect the cardiovascular condition of the body. Measurements to quantify retinal vascular structure and properties have shown to provide good diagnostic capabilities for the risk of cardiovascular diseases. The central retinal artery equivalent (CRAE) and the central retinal vein equivalent (CRVE) are measurements of the diameters of the six largest arteries and veins in the retinal image respectively. These measurements are found to have good correlation with hypertension, coronary heart diseases and stroke. However, they require the accurate extraction of distinct vessels from a retinal image. This is a challenging problem due to ambiguities caused by vessel bifurcations and crossovers. Fig. 1(a) Wrong Identification of I and II Fig. 1(b) Correct Identification of I and II Fig. 1a shows an example retinal image where vessels I and II cross each other at two places (indicated by circles). These crossovers are often mistaken as vessel bifurcations. Fig.c1b shows the correctly identified vessel structure for vessels I and II marked in blue and red respectively. Note that the line segment at the second crossing (larger circle) is shared by vessels I and II. In this paper, a novel technique that utilizes the global information of the segmented vascular structure to correctly identify true vessels in a retinal image. The segmented vascular structure is modeled as a vessel segment graph, and transforms the problem of identifying true vessels to that of finding an optimal forest in the graph. Therefore, an automated identification and separation of individual vessel trees and the subsequent classification into RESEARCH ARTICLE OPEN ACCESS
  2. 2. K. Sureka et al Int. Journal of Engineering Research and Applications www.ijera.com ISSN : 2248-9622, Vol. 4, Issue 3( Version 4), March 2014, pp.01-05 www.ijera.com 2 | P a g e Vessel Segmentation arteries and veins may be requires for vessel specific morphology analysis. II. PROPOSED METHOD An automated method is introduced for structural mapping of retinal vessels by modeling the vessel segmentation into a vessel segment map and identifying the vessel trees based on graph search. Retinal vessel extraction involves segmentation of vascular structure and identification of distinct vessels by linking up segments in the vascular structure to give complete vessels. One branch of work, termed vessel tracking, performs vessel segmentation and identification at the same time. These methods require the start points of vessels to be predetermined. Each vessel is tracked individually by repeatedly finding the next vessel point with a scoring function that considers the pixel intensity and orientation in the vicinity of the current point in the image. Bifurcations and crossovers are detected using some intensity profile. This approach of tracking vessels one at a time does not provide sufficient information for disambiguating vessels at bifurcations and crossovers. Another branch of works treat vessel identification as a post-processing step to segmentation. A graph formulation was used with dijkstra’s shortest path algorithm to identify the central vein. Similarly, Dijkstra’s algorithm used to identify vessels one at a time and evaluated their method on a set of 15 images. However, these methods may lead to incorrect vessel identification because choosing the correct vessel segment to connect at a bifurcation or crossover requires information from other nearby vessels. Our approach differs from existing works in that we identify multiple vessels simultaneously and use global structure information to figure out if linking a vessel segment to one vessel will lead to an overlapping or adjacent vessel being wrongly identified. Fig. 2. Block diagram of proposed method 2.1 Vessel Segmentation and Image Preprocessing The retinal vessels are segmented using the standard approach (supervised pixel classification approach using a Gaussian filter set and classification by a k-nearest neighbor classifier).The binary vessel image is generated from the vessel probability image using Otsu’s thresholding method. The Otsu threshold minimizes the intra-class variance for the foreground (vessel) and background (non-vessel region) classes. Next, the vessel skeleton is obtained by applying mathematical morphology reducing the vessel to a centerline of single pixel width. Fig. 3. (a) Vessel probability (b) Binary image 2.2 Localization of branch points and crossing points The vessel skeletons have to be converted into vessel segments separated by interruptions at the branch and crossing points. Their start and end points are determined by the centerline pixels on the vessel skeleton is analyzes for its 3x3 neighborhood, and branch and crossing points are detected as centerline pixels with more than 2 neighbors. The detection of Preprocessing Localization of branching and crossover Structural separation with Dijkstra’s graph search Identification of AV crossing AV classification using graph cut algorithm Performance evaluation
  3. 3. K. Sureka et al Int. Journal of Engineering Research and Applications www.ijera.com ISSN : 2248-9622, Vol. 4, Issue 3( Version 4), March 2014, pp.01-05 www.ijera.com 3 | P a g e vessel end points is required for the graph search and it is determined as the centerline pixels with only one neighbor. Fig. 4. (a) Vessel network (b) Vessel tree 2.3 Structural separation with Dijkstra’s graph search A vessel consists of number of smaller vessel segments linked together. Three attributes are orientation, width, and intensity of vessel segments corresponding to a single vessel, have similar characteristics within a vessel tree. A vessel subtree is identified by selecting a group of segments from the vessel segment map, based on the similarity between these segments. Three features a) Orientation is expressed as the angle (in radian) the segment end region makes with the positive direction of X-axis, a measurement between [0,π], b) Width (in pixel) is measured as a median value, and passing through the skeleton pixels of the end region, Intensity is measured as a median value of the width and the intensity measured for each vessel segment obtained across the vessel tree. Fig. 5. (a)Vessel segment map (b) Graph structure To convert the vessel segment map into connected graph structure, connecting neighboring vessel segments are identified using the branch and crossing point information. Dijkstra’s algorithm is utilized to identify a vessel subtree. It is a shortest path algorithm to identify the central vein. Similarly, used Dijkstra’s algorithm to identify vessels one at a time and evaluated their method on a set of 15 images. 2.4 Identification of artery-venous crossing I proposed an automated AV separation algorithm based on structural mapping, which classifies the vessel trees into arteries and veins, using vessel color features as well as the anatomic property of arteries-venous (AV) crossing. This property proposes that the crossing of two retinal blood vessels imaged on a two dimensional fundus image, signifies high probability of one vessel being an artery and other one being a vein. The vessel segments are skeletonized to obtain the vessel centerlines. For the centerline extraction, significantly large vessel width segments in a vessel tree are selected to avoid the inclusion of smaller, peripheral or single pixel width segments and is determined as the width more than 60% of the maximum vessel width obtained in that vessel tree. Fig. 6. (a) Vessel probability map (b) Structural mapping A feature vector consisting of four features mean (MG) and standard deviation (SG) of green channel and hue channel respectively, from 3x3 neighborhood of each vessel centerline pixel. Arteries appear brighter (higher green channel intensity) than veins because oxygenated hemoglobin is less absorbent than the de-oxygenated blood between 600-800 nm. 2.5 Artery-venous classification of retinal vessel The centerline pixels obtained from any two vessel trees are collected and classified to detect the AV status of respective vessel trees. Based on feature vector, the algorithm classifies the centerline pixels obtained from a pair of vessel trees, into two clusters/classes. Fig. 7. (a) Structural mapping (b) Artery-venous classification III. RESULTS To evaluate the accuracy of the proposed method, the automated labeling was compared with the expert annotation in terms of a segment color value. Two metrics were utilized to quantify the accuracy of the method. The first metric calculates the mis-classification rate (%) for vessel segments as a function of vessel segment width, over the dataset. The mis-classification rates (%) for various vessel segment sizes were categorized in table. The average mis-classification rate (%) for vessel width above 4
  4. 4. K. Sureka et al Int. Journal of Engineering Research and Applications www.ijera.com ISSN : 2248-9622, Vol. 4, Issue 3( Version 4), March 2014, pp.01-05 www.ijera.com 4 | P a g e pixels was 3.58%. The second metric shows the histogram of pixel mis-classification (%) per image in the dataset. For each image the mis-classification (%) was calculated as the fraction of total number of vessel pixels which was mis-classified, representing its impact on the vessel network. The average mis- classification of 8.56% or the accuracy of 91.44% correctly classified vessel pixels was obtained over the dataset. Fig. 8. (a) Fundus image (b) Structural mapping (c) manual AV labeling (d) Automated AV classification IV. DISCUSSION AND CONCLUSION I developed an automated method for identifying and separating the retinal vessel trees in color fundus images, which provides the mapping of primary vessels, and their branches. The image with highest mis-classification of 44.26% was partially contributed by both false structural mapping and false AV classification. This approach has the potential to impact the diagnostically important morphologic analysis of individual retinal vessels. Vessel size Vessel width Vessel Segment mis-classification (%) Small/Peripheral 1≤width<4 4.07 Medium 4≤width≤6 3.78 Major 6<width≤9 0.00 Table 1: Proportion of mis-classified vessel segments Fig. 9. Quantitative results (a) Proportion of mis-classified vessel segments (b) percentage mis-classification per image
  5. 5. K. Sureka et al Int. Journal of Engineering Research and Applications www.ijera.com ISSN : 2248-9622, Vol. 4, Issue 3( Version 4), March 2014, pp.01-05 www.ijera.com 5 | P a g e REFERENCES [1] Lau Q, Lee M, Hsu W, Wong T (2013) Simultaneously identifying all true vessels from segmented retinal images. IEEE Transactions on Biomedical Engineering 60(7): 1851–58. [2] Rothaus K, Jiang X, Rhiem P (2009) Separation of the retinal vascular graph in arteries and veins based upon structural knowledge. Image and Vision Computing 27(7): 864–875. [3] Joshi V, Garvin M, Reinhardt J, Abramoff M (2011) Identification and reconnection of interrupted vessels in retinal vessel segmentation. In: IEEE, ISBI, Image Segmentation Methods. Volume FR-PS3a.7, pp. 1416–1420. [4] Vickerman M, Keith P, Mckay T Vesgen (2009) 2d: Automated, user-interactive software for quantification and mapping of angiogenic and lymphangiogenic trees and networks. The Anatomical record 292. [5] Kondermann C, Kondermann D, Yan M (2007) Blood vessel classification into arteries and veins in retinal images. In: Med. Imag. Image Process. Volume 6512, p. 651247651249. [6] Abramoff M, Niemeijer M, Suttorp-Schulten M, Viergever M, Russell S, et al.(2008) Evaluation of a system for automatic detection of diabetic retinopathy from color fundus photographs in a large population of patients with diabetes. Diabetes Care 31: 193–198. [7] Witt N, Wong T, Hughes (2006) Abnormalities of retinal microvascular structure and risk of mortality from ischemic heart disease and stroke. Hypertension 47(5): 975–981. [8] Sukkaew L, Makhanov B, Barman S, Panguthipong S (2008) Automatic tortuosity-based retinopathy of prematurity screening system. IEICE transactions on information and systems 12. [9] Koreen S, Gelman R, Martinez-Perez M 2007) Evaluation of a computer-based 399 system for plus disease diagnosis in retinopathy of prematurity. Ophthalmology 114(12): e59–e67. [10] Y. Yin et al., “A probabilistic based method for tracking vessels in retinal images,” in IEEE ICIP, sept. 2010, pp. 4081–4084. [11] T. Y. Wong et al., “Retinal vascular caliber, cardiovascular risk factors, and inflammation: the multi-ethnic study of atherosclerosis (mesa).” Invest Ophthalmol Vis Sci, vol. 47, no. 6, pp. 2341–2350, 2006. [12] H. Li et al., “Automatic grading of retinal vessel caliber,” IEEE Trans. on Biomed. Eng., vol. 52, no. 7, pp. 1352–1355, 2005. [13] J. Cousty et al., “Watershed cuts: Minimum spanning forests and the drop of water principle,” IEEE Trans. on Pattern Anal. And Mach. Intell., vol. 31, no. 8, pp. 1362– 1374, 2009. doing her ME in Applied Electronics at Jayam College of Engineering and Technology, Dharmapuri. Presently she is involving in developing a automated method for identification and classification of retinal blood vessels to identify the diseases in retina. She has published more than two research papers in national and international conferences. Her special areas of interest are Image processing, Control system and Measurements & Instruments. Applied Electronics from Jayam College of Engineering and Technology. He published more than four research papers in various national and international conferences/journals. At present he is working as Assistant Professor in the department of Electronics and Communication Engineering in Jayam College of Engineering and Technology, Dharmapuri. He has participated in various national level workshops and seminars at various colleges. Sureka. K has obtained her BE degree in Electronics and Instrumentation Engineering from Velammal Engineering College, Chennai in 2011. Currently she is Engineering and Vignesh. R has obtained his BE degree in Electronics and Communication Engineering from Jayam College of Engineering and Technology. He received his ME in