643 648

ISSN: 2278 – 1323
International Journal of Advanced Research in Computer Engineering & Technology
Volume 1, Issue 4, June 2012

A Morphological approach for discrimination
between Glaucomatous and Non glaucomatous
Eyes by Optic Disc Localization
C.Radhika,radhika.scn@gmail.com
TamilNadu, India.

The increase in pressure damages the optic nerve that carries
Abstract-A Portion of optic nerve that is visible in the retinal the vital information from the retina to the brain. The damage
fundus is called the optic disc (OD).Optic disc is one of the vital can be avoided only by treatment that can reduce or prevent
information in detecting the diseases in the retina by further damages. Visually, the damage is observed in the
implementing an image processing algorithm. Algorithm focuses relative areas of the optic disc and the cup within the disc.
on the properties and the characteristics of the optic disc. Optic The normal optic disc is the site of passage from the eye of
disc has got its own shape, color, size and intensity used to detect more than 1 million axons that connect their retinal ganglion
it in the retinal images. The optic disc is the entrance of the optic cells. The diameter of the normal optic disc varies from eye to
nerve and the vessels in the retina called blind spot. Blind spot in
eye. As the nerve fibers die in the patients who were affected
the retina has the cone cells, rod cells which captures the
impulses and convert them into proper images and carry them to
by glaucoma, the diameter of the cup within the optic disc gets
the brain. Detection of optic disc is useful in diagnosing the enlarged. The normal cup to disc ratio is about 0.3 mm,
pressure within the skull that causes this disc to bulge forward. whereas with the glaucoma injury it is increased to more than
The pressure within the eye squeezes off the blood supply, and 0.7 mm. Abnormality detection in retinal fundus images is
reduced blood supply starves the retina from nourishment, and predicted to play an important role in many real life
ultimately the nerve cells start to die - glaucoma. A novel applications. A fast, accurate and reliable method for
approach is proposed to investigate the screening performance on abnormality detection in images will help greatly in improving
the glaucoma disease, and to differentiate the normal and the health care screening process.
glaucomatous eyes, this screening system is designed. With Cup The retinal fundus photographs are widely used in the
to disc ratio (C/D), the classification among the normal and the
diagnosis and treatment of various eye diseases in clinics. It is
glaucomatous images was found. When the cup to disc is greater
than 0.3 (cup to disc> 0.3mm), it is considered as the glaucoma also one of the main resources for mass screening of
affected eye. glaucoma. Being able to automatically and quickly process, a
large number of fundus images can help the ophthalmologists
increase the productivity and efficiency in medical field.
Index Terms— Glaucoma, cup to disc ratio, optic disc
Developing an automatic fundus [14] image analyzing and
detection, interface tool, screening system.
diagnosing system has been the ultimate aim to facilitate the
I. INTRODUCTION clinical diagnosis.
Extraction of the normal and abnormal features in color
Detection of optic disc is an indispensable step in the fundus images is the fundamental method that is useful for
automatic analysis of digital color fundus images. Optic disc automatic understanding of fundus images. It is concerned
detection is an integral part of the screening system for with developing automated techniques of generating
glaucoma. The retina contains anatomic structures like the quantitative descriptions of the retina that might be used in
vasculature, optic disc, and macula, when the system has diagnosis and treatment. The automatic determination of the
located these structures in the image then, these locations will position of distinctive points is an important and intermittent
be used as the landmarks for the detection of the symptoms for theme in medical image analysis. A wholly automated
glaucoma. The optic nerve is the region of the retina where the approach involving computer analysis of fundus images could
blood vessels and nerve fibers pass through the sclera. It is provide an immediate classification of glaucoma without the
sensitive to the changes associated with intraocular pressure specialist opinions.
associated with glaucoma that may occur asymptomatically This article describes about the optic disc location and
and which can be diagnostic and that must be tracked to glaucoma screening system using various methods to pre-
monitor the progress of treatment. process the image to improve the quality and computer based
One of the symptoms of glaucoma is an increase in algorithms were employed for the detection of optic disc in the
pressure within the eye. It is caused as a result of blockage of retinal image, and various techniques were used to recognize
the flow of aqueous humour, a watery fluid produced by the the important features of glaucoma.
ciliary body.

643
All Rights Reserved © 2012 IJARCET

ISSN: 2278 – 1323

II. LITERATURE SURVEY technique, the algorithm starts with a manually-entered initial
In the literature a number of different techniques have been point and an initial direction, and recursively tracks the entire
arterial tree using a breadth-first search. This would not be useful
employed to automatically detect the optic disc. In general
for retinal images since the vessels are not necessarily connected,
these techniques apply preprocessing operations, followed by
especially in partial views.
several image processing operations and finally the OD-
localization. A preprocessing is for instance the generation of
III METHOD
a mask image to determine which area belongs to the actual
fundus and which area belongs to the background of the A Overall Scheme
image. Some of the automatic OD- detection methods produce Our goal is to detect the blind spot of the retina called optic
only a point that can be used as the ODcenter. disc and to identify whether the retina is affected by glaucoma
In [1] Osareh propose a template-based OD-detection. This disease or not. Many algorithms were developed for detecting the
method uses a color normalization of the fundus image. The optic disc in the retinal fundus images. Many algorithms were
template-based OD-detection method assumes the optic disc to theoretically good and many methodologies concentrates on the
be approximately circular and consisting of bright pixels. At optic disc detection alone. Hence the good interface for disease
first the color images are normalized and then the intensity detection has not been proposed. The proposed system is being
components from the HSI space are used to create a template the interface between the optic disc detection and the screening
and to perform the template matching. The normalization of system for the glaucoma disease. The following are the various
the color fundus images is performed by applying histogram steps involved in detecting the optic disc and identifying the
specification on each color plane (R, G and B). Histogram disease in the retinal image. (1)Color conversion for intensity
specification requires one image specifying the preferred enhancement, (2) grayscale conversion, (3) fast level for multi
histogram of the color plane. The histogram of this image is level thresholding, (4) optic disc identification, (5) morphological
used to approximate the new histogram of the image to operations, (6) optic disc circle measurement, (7) identification
normalize, this way the appearance of the normalized image of cup to disc ratio, and (8) classification between normal and
approaches the appearance of the model image. glaucomatous eyes.
In [2], Walter proposes an OD-localization method that
applies a threshold to obtain pixels with high intensity values B. Color Conversion for Intensity Enhancement
and selects the center of the largest object as the OD-center. Color conversion process is the initial process in the image
The detection of the optic disc is performed on the intensity processing. This process facilitates thresholding and used to blur
component from the HSI space. In the intensity image the the noise in the image and to smooth the boundaries of the
optic disc is assumed to be the largest brightest part of the particles to be detected. Pre-processing methods use a small
image. A simple threshold is applied to obtain a binary image neighborhood of a pixel in an input image to get a new brightness
containing parts of the optic disc and perhaps other bright value in the output image. Such pre-processing operations are
appearing pathologies like exudates. The largest connected called also filtration. The image generally will be highly
object within the threshold image is expected to be a part of correlated; most of the colors are muted. To overcome this
the optic disc. The center of this object is selected as the center
drabness, it is necessary to get more pure colors. So color
of the optic disc.
manipulation is very important as shown in Fig-1.
In [3], Barrett proposes a method based on a Hough
transform in order to locate the optic disc. The Hough
transform technique is able to find geometric shapes in an
image. Objects of geometric shapes may be detected by
converting the equation of the object into a Hough space
parameter equation.
In [4], optic disc is often a bright circular shape at the
convergence of the vasculature. This method assumes that the
OD -center lies close to a vessel of the vasculature. However,
this time the Hough transform is only applied on and close to Fig 1: Color conversion for intensity enhancement
the vasculature. In order to determine the potential OD-
locations the segmentation of the vasculature is required. C. Grayscale Conversion
In [5] three exploratory processing techniques are described. The gray scale image should visibly reflect the magnitude of
In the first technique, commonly used in quantitative coronary the color contrast. The transformation from color to grayscale is a
analysis (QCA), the initial and end points of the vessel continuous function. Grayscale is a range of shades of gray
(sometimes also the direction and width) are entered manually. without apparent color. The darkest possible shade is black,
Although these algorithms are very accurate, they are which is the total absence of transmitted or reflected light This
unsuitable for real-time retinal image processing since they constraint reduces image artifacts, such as false contours in
require manual input and suffer from high computational time, homogeneous image regions. When a pixel in the color image is
which are not compelling constraints in QCA. In the second gray, it will have the same gray level in the grayscale image. This

644

ISSN: 2278 – 1323

constraint assists in image interpretation by enforcing the (σ B)^2 = ω 1(µ 1- µ t) ^2 + ω 2(µ 2- µ t) ^2 Eq.9
usual relationship between gray level and luminance value. Otsu verified that the optimal threshold t* is chosen so that the
Grayscale is best because it produces the result whose between class variance is maximized. That is
brightness is the most perceptually similar to the brightness of t* = Max {(σ B) ^2 (t)} Eq.10
the original color image – Fig-2.
E. Optic Disc Identification
Once the threshold value is identified using the fast level multi
thresholding, the region of interest is identified .The region less
than the threshold value is considered as the part1 and the region
greater than the threshold value is considered as the part2. Part1
consist of the other region such as blood vessel, fovea, macula
etc. Part 2 consist of the optic disc which is the landmark for
identifying the disease called glaucoma. In the abnormal images
the features of the optic disc vary from the normal retinal images.
Hence the features of optic disc such as bright yellowish region,
Fig 2: Gray Retinal Fundus Image the shape, and the size of the optic disc will not help to identify
the land mark region. Hence the fast level multi thresholding is
D. Fast level for multilevel thresholding adopted to identify the optic disc both in normal and abnormal
Thresholding is an important technique for image images.
segmentation that tries to identify and extract a target from its
background on the basis of the distribution of gray levels or F. Morphological Operations
texture in image objects. Otsu’s method [7] was one of the Mathematical morphology can extract important shape
better threshold selection methods for general real world characteristics and also remove irrelevant information. Using
images with regard to Uniformity and shape measures. An grey level morphology, the operation can be applied to the
image is a 2D grayscale intensity function, and contains N intensity or lightness channel. The best method of obtaining a
pixels with gray levels from 1 to L. In the case of bi- level homogeneous OD region by performing grey level morphology.
thresholding of an image, the pixels are divided into two Opening and closing are the morphological operators. Opening
classes, C1 with gray levels [1… t] and C2 with gray levels smoothes the contour of an object, and eliminates the thin
[t+1… L]. the gray level probability distribution for the two protrusions. Closing tends to smooth sections of contours but,
classes is found and the means for the two classes are also unlike opening, it fuses narrow breaks and eliminates the small
found. From the means of the two classes, the mean intensity hole, and fills the gap in the contour. The closing operation is
for the entire image is found. Using discriminant analysis, performed i.e. a dilation to first remove the blood vessels and
Otsu defined the between-class variance of the threshold then an erosion to restore the boundaries to their former position.
image. For bi-level thresholding, Otsu verified that the optimal This can result in some minor inaccuracies, particularly if any
threshold is chosen so that the between-class variance is boundary concavities are filled by the dilation, but in the main
maximized and the maximum variance is chosen as the performs very well. The other operation is opening operation is
optimal threshold (Fig-3a). performed i.e. a erosion which is similar to dilation but opposite
Optimal Thresholding Methodology: (OTM) effect. It removes the pixels from the edges of objects within an
An image is 2Dimensional (2D) gray images and it contains image.
N pixels with the gray levels 1 to L. The probability of gray Closing Operation
level i in an image is Closing is an important operator from the field of mathematical
pi=fi/N. Eq.1 morphology. Like its dual operator opening, it can be derived
The gray level’s probability distributions for the two classes from the fundamental operations of erosion and dilation. Like
C1 and C2 are those operators it is normally applied to binary images, although
C1 = p1/w1(t) …. Pt/w1(t) Eq.2 there are gray level versions. Closing is similar in some ways to
C2 = pt+1/w2(t)…… pL/w2(t) Eq.3 dilation in that it tends to enlarge the boundaries of foreground
(bright) regions in an image (and shrink background color holes
ω 1(t) = ∑ρi (for i=1 to t) Eq.4 in such regions), but it is less destructive of the original boundary
ω2(t) = ∑ρi (for i=1 to t+1) Eq.5 shape. As with other morphological operators the exact operation
is determined by a structuring element. The effect of the operator
Means for the classes C1 and C2 are is to preserve background regions that have a similar shape to
µ 1 = ∑ i ρi/ ω 1(t) (for i= 1 to t) Eq.6 this structuring element, or that can completely contain the
µ 2 = ∑ i ρi/ ω 2(t) (for i= t+1 to L) Eq.7 structuring element, while eliminating all other regions of
Let µ t be the mean intensity for the entire image. background pixels. This operation uses the two functions
Ω1µ1+ ω2µ2=µt Eq.8 structuring element T and the input fundus image S
Otsu defined the between class variance of the thresholded
image (S+T) – T Eq.11

645

ISSN: 2278 – 1323

(2) Y- the y-co-ordinate of the center of the circle
Opening Operation (3) R- the radius of the circle
Opening is the dual of closing, i.e. opening the foreground
pixels with a particular structuring element is equivalent to
closing the background pixels with the same element. The
effect of opening can be quite easily visualized. Imagine
taking the structuring element and sliding it around inside each
foreground region, without changing its orientation. All pixels
which can be covered by the structuring element with the
structuring element being entirely within the foreground
region will be preserved. However, all foreground pixels
which cannot be reached by the structuring element without
parts of it moving out of the foreground region will be eroded Fig 3 (a) Optic disc of the Retinal Fundus Image after
away. After the opening has been carried out, the new Morphological Operation
boundaries of foreground regions will all be such that the
structuring element fits inside them, and so further openings
with the same element have no effect.
This operation uses the two functions structuring element T
and the input fundus image S

(S - T) + T Eq.12

G. Structuring Element
All these morphological operators take two pieces of data as
input. One is the input image, which may be either binary or
grayscale for most of the operators. The other is the
structuring element. It is this that determines the precise
Fig 3(b) Circle fitting Operation on retinal fundus image
details of the effect of the operator on the image. The
structuring element consists of a pattern specified as the
I. Cup to Disc Ratio (C/D Ratio)
coordinates of a number of discrete points relative to some
origin. More complicated elements, such as those used with The optic nerve carries impulses for sight from the retina in
thinning or grayscale morphological operations may have the eye to the brain. It is composed of millions of retinal nerve
other pixel values. When a morphological operation is carried fibers that bundle together and exit to the brain through the optic
out, the origin of the structuring element is typically translated disc located at the back of the eye. The optic disc has a center
to each pixel position in the image in turn, and then the points portion called the ―cup‖ which is normally quite small in
within the translated structuring element are compared with comparison to the entire optic disc. In people with glaucoma
the underlying image pixel values. The details of this damage, because of increased pressure in the eye and/or loss of
comparison and the effect of the outcome depend on which blood flow to the optic nerve, these nerve fibers begin to die.
morphological operator is being used. This causes the cup to become larger in comparison to the optic
disc, since the support structure is not there. A cup to disc ratio
H. Optic Disc Circle Measurement greater than six- tenths is generally considered to be suspicious
Fitting circles to given points in the plane is a problem that for glaucoma. Through periodic photographs of the optic nerve,
arises in many application areas, e.g. computer graphics, the ratio of the cup to the disc can be monitored. Glaucoma
medical image processing, statistics. Here the least square affects the optic nerve head causing cupping and nerve cell fibers
fitting algorithm is used since the shape of the optic disc is are destroyed. This destruction of healthy nerve fiber cells at the
considered as the circle, hence to fit the circle around the optic optic nerve causes loss of the peripheral visual field. This
disc region this algorithm is used. There are several algorithms cupping is the hallmark sign of glaucoma. A cup -to-disc ratio [5]
such as ellipse square fitting algorithm, but it won’t suite for is critical when evaluating glaucoma. The cup-to-disc ratio is the
the optic disc. Only when the shape of the optic disc is amount of the entire nerve head that has been cupped out or
considered as the circle, the radius, diameter, and the cup to where glaucoma has caused damage. Using these optic disc
disc ratio of the optic disc can be identified and the measurements, the cup to disc ratio can be calculated. The
classification among the images can be made in order to normal cup to disc ratio i.e. the diameter of the cup divided by
identify the normal and the glaucomatous retinal images the diameter of the whole optic disc is about 0.3 mm. If the ratio
(Fig3b). varies and if it is larger than 0.3mm then it is suspected that the
The parameters used for fitting the circle around the optic disc cup could be getting enlarged. Glaucoma is one of the diseases
are as follows; which is related to the cup to disc ratio parameter. This parameter
(1) X- the x-co-ordinate of the center of the circle is helpful in identifying this disease because the glaucoma can

646

ISSN: 2278 – 1323

cause the cup to enlarge. Among 100 images, 92 images shown the detected optic disc
is exactly the region of interest or some other land marks

Total Sensitivity Specificity Conformity
images
100 95 89 92

Fig 4: The retinal images affected by the disease glaucoma Table 1- Parameters under Study
and the variation in the cup to disc
V CONCLUSION
IV RESULT We have presented the method for locating the optic disc in
The results of the present study lead to a resurgence of the retinal fundus images and to develop methods for separating
cup-to-disc diameter ratios in the clinical diagnosis of normal from abnormal images (cases of glaucoma). These
glaucoma if the dependence of the cup-to-disc diameter ratios would be used in a screening clinic to identify at-risk patients.
on the disc size is taken into account. The optic disc size was Images were collected from various sources and the data
found to be useful clinically, especially to assist in identifying collected at a range of sites. Methods were developed to
small glaucomatous optic discs (Fig-4). separate the normal from the abnormal images. This was done
In the affected eye, the frequency increases with larger with reasonable success. Whilst the modest success could be
cup/disc ratios, being greatest for C/D of 0.8 to 0.9, whereas in attributed to the insensitivity of the analysis, it can also be
the control, the frequency is highest for C/D values between attributed to the nature of the diagnosis. we are labeling
0.0 and 0.3 and decreases markedly and progressively for images as being abnormal or not, without recognizing that
larger C/D values so that the least frequent is 0.8 to 0.9. In the there is a spectrum of appearances. The tests indicate that the
unaffected eye, the frequency distribution of C/D ratio is also optic disc’s appearance is more uniform in the normal and
different from the control group. Frequency of C/D of 0.0 to becomes progressively less so as the diseases progress. The
0.1 is very small compared to the control, whereas that of screening system acts as an interface tool for the early
values greater than 0.3 is comparatively greater. These detection of the glaucoma disease and also could serve as one
differences in frequency are statistically significant at the 1 per of the module in the medical diagnosing system in the medical
cent level of confidence. field.

Parameters under study VI FUTURE ENHANCEMENT
The fundus images from various datasets were collected and Future work can focus on two directions: accumulating further
tested against the proposed methodology. Among those data with respect to the variables for the optic disc as well as
images, 100 images were included and analyzed. These 100 for detecting the various other diseases related to the optic disc
images worked exactly against the proposed methodology and developing more robust and accurate methods of
under three parameters namely sensitivity, specificity, and processing the screening system in the clinical environment.
conformity as shown in Table-1. The modification can be done in such a way that the system
a) Sensitivity: It specifies the classification among the normal should be able to recognize various images from various
and glaucomatous retinal fundus images datasets and the screening system can be used in the clinical
b) Specificity: It specifies whether the region of interest i.e. environment at large extent. The diabetic retinopathy is also
the optic disc is exactly normal or affected by the disease another disease which affects the optic disc region in the
called glaucoma retinal fundus image, hence this screening system can be
c) Conformity: It specifies the detected optic disc is correct or widen for screening this disease and this interface tool will be
some other landmark other than the optic disc. one of the module in medical diagnosing field in the medical
. environment.
Sensitivity
Among 100 images, 95 images were correctly sensed and REFERENCES
the classification was among normal and the glaucoma [1] A. Osareh, M. Mirmehdi, B. Thomas, and R. Markham. Comparison of
affected images. colour spaces for optic disc localisation in retinal images. Proceedings of
the 16th International Conference on Pattern Recognition, pages 743–
Specificity 746, 2002.
Among 100 images, 89 images were perfectly specified the [2] T. Walter and J.C. Klein. Segmentation of color fundus images of the
human retina: Detection of the optic disc and the vascular tree using
region of interest i.e. the optic disc. In these 13 images the morphological techniques. Proceedings of the 2nd International
optic disc were correctly detected both in normal and in the Symposium on Medical Data Analysis, pages 282–287, 2001.
glaucoma affected images. [3] S.F. Barrett, E. Naess, and T. Molvik. Employing the Hough transform
to locate the optic disk. Biomedical Sciences Instrumentation, 37:81–86,
Conformity 2001.

647

ISSN: 2278 – 1323

[4] M. Niemeijer, J.J. Staal, B. van Ginneken, M. Loog, and M.D.
Abramoff. Comparative study of retinal vessel segmentation methods
on a new publicly available database. SPIE Medical Imaging,
5370:648– 656, 2004.
[5] A.Can, H. Shen, J. N. Turner, H. L. Tanenbaum, and B. Roysam,
―Rapid automated tracing and feature extraction from live high-
resolution retinal fundus images using direct exploratory algorithms,‖
IEEE Trans. on Info. Tech. for Biomedicine, 3(2):125– 138, 1999.
[6] Grehn, Franz MD, ―World Health Problem of Glaucoma,‖ Journal of
Glaucoma. 10(5) Supplement 1:S2-S4, October 2001.
[7] Pederson JE, Anderson DR. The mode of progressive disc cupping in
ocular hypertension and glaucoma. Arch Ophthalmol 1980;98:490–5.
[8] Ping-Sung Liao, Tse-Sheng Chen* And Pau-Choo Chung, A Fast
Algorithm for Multilevel Thresholding Department of Electrical
Engineering ChengShiu Institute of Technology Kaohsiung,
Department of Engineering Science National Cheng Kung University
Tainan Taiwan, JOURNAL OF INFORMATION SCIENCE AND
ENGINEERING 17, 713-727 (2001) 713
[9] Chaudhuri, N. K. Chatterjee, and M. Goldbaum, ―Automatic Detection
of the Optic Nerve in Retinal Images‖, presented at Proceedings IEEE
International Conference on Image Processing, Singapore, pp. 1-5,
1989.
[10] Sommer A, Pollack I, Maumenee AE. Optic disc parameters and onset
of glaucomatous visual field loss, I: methods and progressive changes
in disc morphology. Arch Ophthalmol. 1979;97:1444–1448.
[11] R.Harper, Barnaby Reeves, ―The sensitivity and specificity of direct
ophthalmoscopic optic disc assessment in screening for glaucoma: a
multivariate analysis,‖ Graefe's Archive for Clinical and Experimental
Ophthalmology, Volume 238, Number 12,pp. 949 – 955, December
2000
[12] Quigley HA, Addicks EM, Green WR, Maumenee AE. Optic nerve
damage in human glaucoma, III: quantitative correlation of nerve fiber
loss and visual field defect in glaucoma, ischemic optic neuropathy,
papilledema and toxic neuropathy. Arch Ophthalmol.1989;107:453–
464.
[13] H. Li and O. Chutatape, ―Automatic Location of Optic Disk in Retinal
Images‖, presented at Image processing, Thessaloniki, Greece, pp. 837-
840, Oct, 2001.
[14] I.Constable, M. McCombe, P. Mitchell, J. O'Day, P. Phillips, A.
Stocks, H.Taylor, and T. Welborn, Diabetes and the Eye - Professional
Guidelines: Australian Diabetes Society for Diabetes Australia, 1996.
[15] Michelson, Georg MD, Groh, Michael J.M. MD, ―Screening models
for glaucoma.‖ Current Opinion in Ophthalmology, 12(2):105-111,
April 2001.
nd
[16] Gonzalez and Woods., ―Digital Image Processing 2 Edition‖,
Prentice Hall, Copyright 2002.

C.Radhika received B.E degree in Electrical and
Electronic Engineering and M.Tech Degree in
Information Technology from Vellore Institute
of technology. Research Interest includes Mobile
communication, Wireless sensor Network, And Imageprocessing .

648

643 648

More Related Content

What's hot (20)

Viewers also liked (7)

Similar to 643 648 (20)

More from Editor IJARCET (20)

Recently uploaded (20)

643 648