Characterization of changes in blood vessel width and tortuosity in retinopathy of prematuriIntroductionMathematical preliminariesNotationExtraction and characterization of vessel widths and tortuosityCharacteristics of the vasculatureExisting work in vasculature segmentationMatched filteringMorphological methodsArtificial neural networksVessel tracking algorithmsManual techniquesProposed approach to segmentationInitial morphological filteringSecond derivative properties of the vasculatureFinal morphological filteringThresholdingFurther processingThe skeletonisation and skeletal filtering algorithmIdentification of fiducialsFinding the vessel widthCalculation of vessel tortuosityImplementationResultsAccuracy of segmentation and parameter sensitivityAccuracy of vessel width and tortuosity measurementAnalysis of ROP imagesPotential implications for automated screeningConclusions and discussionAcknowledgementsReferencesMedical Image Analysis 6 (2002) 407–429www.elsevier.com/locate/mediaC haracterization of changes in blood vessel width and tortuosity inretinopathy of prematurity using image analysisa, a,b c c,d*Conor Heneghan , John Flynn , Michael O’Keefe , Mark CahillaDepartment of Electronic and Electrical Engineering,University College Dublin,Belfield,Dublin4,IrelandbDepartment of Electrical Engineering and Computer Sciences,University of California,Berkeley,CA94720,USAcDepartment of Ophthalmology,The Children’s Hospital,Temple Street,Dublin1,IrelanddBeetham Eye Institute,Joslin Diabetes Center,1Joslin Place,Boston MA02215,USAReceived 23 March 2001; received in revised form 1 August 2001; accepted 22 October 2001AbstractMany retinal diseases are characterised by changes to retinal vessels. For example, a common condition associated with retinopathy ofprematurity (ROP) is so-called plus disease, characterised by increased vascular dilation and tortuosity. This paper presents a generaltechnique for segmenting out vascular structures in retinal images, and characterising the segmented blood vessels. The segmentationtechnique consists of several steps. Morphological preprocessing is used to emphasise linear structures such as vessels. A secondderivative operator is used to further emphasise thin vascular structures, and is followed by a final morphological filtering stage.Thresholding of this image is used to provide a segmented vascular mask. Skeletonisation of this mask allows identification of points inthe image where vessels cross (bifurcations and crossing points) and allows the width and tortuosity of vessel segments to be calculated.The accuracy of the segmentation stage is quite dependent on the parameters used, particularly at the thresholding stage. However,reliable measurements of vessel width and tortuosity were shown using test images. Using these tools, a set of images drawn from 23subjects being screened for the presence of threshold ROP disease is considered. Of these subjects, 11 subsequently required treatment forROP, 9 had no evidence of ROP, and 3 had spontaneously regressed ROP. The average vessel width and tortuosity for the treated subjectswas 96.8 m m and 1.125. The corresponding figures for the non-treated cohort were 86.4 mm and 1.097. These differences werestatistically significant at the 99% and 95% significance level, respectively. Subjects who progressed to threshold disease during thecourse of screening showed an average increase in vessel width of 9.6 mm and in tortuosity of 10.008. Only the change in width wasstatistically significant. Applying a simple retrospective screening paradigm based solely on vessel width and tortuosity yields a screeningtest with a sensitivity and specificity of 82% and 75%. Factors confounding a more accurate test include poor image quality, inaccuraciesin vessel segmentation, inaccuracies in measurement of vessel width and tortuosity, and limitations inherent in screening based solely onexamination of the posterior pole. 2002 Elsevier Science B.V. All rights reserved.Keywords:Morphological processing; Retinopathy of prematurity; Segmentation; Screening1 . Introduction tinopathy is often characterised by the presence of newblood vessels, venous beading, microaneurysms, and intra-Pathological changes of the retinal vasculature are a retinal macular abnormalities. Another example is thefeature of many diseases. For example, diabetic re- presence of plus disease, concurrent with retinopathy ofprematurity, which is characterised by an increase in vesselwidth and tortuosity. Currently, these systematic changes*Corresponding author. Tel.: 1353-1-706-1925; fax: 1353-1-283-in vessel characteristics are determined qualitatively by0921.E-mail address:[email protected] (C. Heneghan). direct inspection (ophthalmoscopy) or through examination1361-8415/02/$ – see front matter  2002 Elsevier Science B.V. All rights reserved.PII: S1361-8415(02)00058-0408 C.Heneghan et al./ Medical Image Analysis6 (2002) 407–429of photographic records of the retina. Relatively little work dilated and tortuous. All the parameters are cumulative andhas been carried out to date on automated quantitative if they are all present in sufficient severity an eye is said toanalysis of digital retinal imagery. This is probably for two have threshold disease. Other inflammatory changes ac-reasons: (a) until recently, digital images of the retina company these vasculature changes and threshold diseasecould only be obtained through scanning of conventional is associated with the imminent development of severe,film photography or slides, which made image processing sight-threatening ROP. Fortunately, the majority of casesrelatively cumbersome, and (b) the tasks required of of ROP regress spontaneously, but once threshold diseaseautomated analysis are quite demanding from an image has been detected treatment is indicated. The mainstay ofprocessing point of view. treatment for ROP is ablation of the immature retina that isIn this paper, we present a technique for automatically producing the abnormal growth factors that stimulate thesegmenting out the retinal vasculature from an image of new blood vessels, using either cryotherapy (Cryotherapythe retina, and for using this segmentation to provide for Retinopathy of Prematurity Co-operative Group, 1988,measurements of vessel widths and tortuosity. The clinical 1990a,b) or laser treatment (McNamara et al., 1992).motivation for undertaking this study was to quantitatively Screening premature infants