Characterization of changes in blood vessel width and tortuosity in retinopathy of prematurity using image analysis

Abstract

Many retinal diseases are characterised by changes to retinal vessels. For example, a common condition associated with retinopathy of prematurity (ROP) is so-called plus disease, characterised by increased vascular dilation and tortuosity. This paper presents a general technique for segmenting out vascular structures in retinal images, and characterising the segmented blood vessels. The segmentation technique consists of several steps. Morphological preprocessing is used to emphasise linear structures such as vessels. A second derivative 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 in the 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 23 subjects being screened for the presence of threshold ROP disease is considered. Of these subjects, 11 subsequently required treatment for ROP, 9 had no evidence of ROP, and 3 had spontaneously regressed ROP. The average vessel width and tortuosity for the treated subjects was 96.8 μm and 1.125. The corresponding figures for the non-treated cohort were 86.4 μm and 1.097. These differences were statistically significant at the 99% and 95% significance level, respectively. Subjects who progressed to threshold disease during the course of screening showed an average increase in vessel width of 9.6 μm and in tortuosity of +0.008. Only the change in width was statistically significant. Applying a simple retrospective screening paradigm based solely on vessel width and tortuosity yields a screening test with a sensitivity and specificity of 82% and 75%. Factors confounding a more accurate test include poor image quality, inaccuracies in vessel segmentation, inaccuracies in measurement of vessel width and tortuosity, and limitations inherent in screening based solely on examination of the posterior pole.

Introduction

Pathological changes of the retinal vasculature are a feature of many diseases. For example, diabetic retinopathy is often characterised by the presence of new blood vessels, venous beading, microaneurysms, and intra-retinal macular abnormalities. Another example is the presence of plus disease, concurrent with retinopathy of prematurity, which is characterised by an increase in vessel width and tortuosity. Currently, these systematic changes in vessel characteristics are determined qualitatively by direct inspection (ophthalmoscopy) or through examination of photographic records of the retina. Relatively little work has been carried out to date on automated quantitative analysis of digital retinal imagery. This is probably for two reasons: (a) until recently, digital images of the retina could only be obtained through scanning of conventional film photography or slides, which made image processing relatively cumbersome, and (b) the tasks required of automated analysis are quite demanding from an image processing point of view.

In this paper, we present a technique for automatically segmenting out the retinal vasculature from an image of the retina, and for using this segmentation to provide measurements of vessel widths and tortuosity. The clinical motivation for undertaking this study was to quantitatively assess a set of clinical imagery obtained from subjects who were being screened for retinopathy of prematurity (ROP). In ROP, normal retinal vessel development is halted and replaced by abnormal blood vessels. It is a serious complication of premature birth with an incidence of between 16 and 56% in infants weighing 1500 g or less or born at a gestational age of 30 weeks or less (McNamara and Connolly, 1999, Palmer et al., 1991), and can lead to severe long-term vision loss or blindness if not treated correctly.

An international classification of ROP has been developed to standardize evaluation and assist in clinical research into the disease (Committee for the Classification of Retinopathy, 1984). The classification divides the eye into a set of zones (Fig. 1) and the severity of ROP is determined according to three parameters: (a) the zone in which new vessels are located, (b) how much of the retina is involved (which is determined by dividing the retinal area into clock hours) and (c) how much fibrosis is associated with the vessels. A fourth independent parameter is the presence or absence of ‘plus disease’. Plus disease is when normal blood vessels located near the optic disc (which is near the posterior pole of the retina) become dilated and tortuous. All the parameters are cumulative and if they are all present in sufficient severity an eye is said to have threshold disease. Other inflammatory changes accompany these vasculature changes and threshold disease is associated with the imminent development of severe, sight-threatening ROP. Fortunately, the majority of cases of ROP regress spontaneously, but once threshold disease has been detected treatment is indicated. The mainstay of treatment for ROP is ablation of the immature retina that is producing the abnormal growth factors that stimulate the new blood vessels, using either cryotherapy (Cryotherapy for Retinopathy of Prematurity Cooperative Group, 1988, Cryotherapy for Retinopathy of Prematurity Cooperative Group, 1990a, Cryotherapy for Retinopathy of Prematurity Cooperative Group, 1990b) or laser treatment (McNamara et al., 1992).

Screening premature infants for threshold disease makes sense from a public health perspective. ROP is a significant health problem, its natural history is known, the at-risk population can be identified, the screening test is sensitive and specific, and an economic treatment that is effective in altering the course of the disease is available. International guidelines have been established for screening. A key issue is to avoid screening before the disease is normally present, which would result in the false impression that there is no ROP, but not delaying too long relative to the natural history of the disease, in case any ROP may be too advanced to respond to treatment. In general, the first examination takes place between 4 and 6 weeks of chronological age or between 31 and 33 weeks post-conceptional age. Subsequent examinations depend on the presence and severity of any ROP detected. Subjects with no ROP may have only one further examination 6 weeks later, whereas subjects with significant disease may require weekly or even daily examinations, to assess if they have reached threshold disease. Typically these examinations are carried out with an indirect ophthalmoscope, a device that allows the ophthalmologist to get a stereoscopic wide-angle view of the retina, but which is quite challenging to use correctly.

The main goal of this screening process is to reliably identify subjects who have progressed to threshold disease, so they can be promptly treated. At present this screening process is carried out by ophthalmologists skilled in the examination of infants’ eyes. Any system which can assist ophthalmologists in increasing the accuracy of their screening, or which could allow less highly trained individuals to carry out the screening (e.g., ophthalmic nurses) may be of clinical benefit. A possibility of providing some automated assistance in this screening process lies in accurate computer measurement of vessel width and tortuosity near the posterior pole (back) of the retina. The clinical justification for this is that it has been recently demonstrated that the absence of dilated and tortuous vessels in the posterior pole is a reliable marker for the absence of threshold ROP (Saunders et al., 2000, Wallace et al., 2000), lessening the need for indirect ophthalmoscopy of the peripheral retina. The posterior pole can be more easily visualised using a direct ophthalmoscope (which is considerably easier to use), or a fundus camera. Therefore analysis of the region near the posterior pole can be used as a screening test in its own right, since only subjects with changes in this region will exhibit threshold disease. This opens the possibility of screening for ROP by non-ophthalmologists using a direct ophthalmoscope (Saunders et al., 2000), or by automated techniques which provide quantitative measurements of vessel width and tortuosity in the posterior pole.

This paper proposes a technique for measuring vessel width and tortuosity in the posterior pole of the retina, and shows how it can be used to demonstrate statistically significant changes in these parameters for ROP subjects. The technique firstly implements segmentation of the retinal vasculature. This is carried out using morphological pre-processing based on linear structuring elements, followed by enhancement using a smoothed second-derivative operator. A final stage of morphological post-processing is then used, prior to thresholding. This results in a binary mask representing detected vessels from which vessel widths can be calculated. This binary mask is then skeletonised. The tortuosity of segments of this skeleton (corresponding to specific blood vessels) can be calculated. By considering the average vessel width and tortuosity at the posterior pole, it can be shown that (a) subjects with threshold disease have wider and more tortuous vessels, though only the changes in vessel width can be considered to be statistically significant based on our measurement technique, (b) that subjects who progress to threshold disease subsequent to their first examination experience a statistically significant increase in average vessel width, and (c) that treatment leads to a rapid reduction in average vessel width and tortuosity. A rudimentary automated screening system for threshold disease based solely on measurement of vessel width and tortuosity is proposed, and the sensitivity and specificity of this test are shown to be 82% and 75%, respectively, for the test data set.