Diabetic retinopathy is a severe vision complication of
diabetes and it is one of the main causes of blindness all over the
world. Early detection of diabetic retinopathy is essential to cope
with this adverse effect. Microaneurysm is one of the early signs
of diabetic retinopathy. So the presence of microaneurysm
detection is a prerequisite for early diagnosis of diabetic
retinopathy. In this paper, we have proposed a simple
morphological method for the detection of microaneurysm that
uses top-hat transform. Our method can detect the faint
microaneurysm at low resolution due to contrast enhancement
and noise reduction as preprocessing. We also compare the
results of different retinopathy detection techniques.
Detection of Microaneurysm in Diabetic Retinopathy
1. Detection of Microaneurysm in Diabetic Retinopathy
Morium Akter
Department of Computer Science and Engineering
Jahangirnagar University
Savar, Dhaka, Bangladesh
e-mail: ecs_morium@yahoo.com
Abstract—Diabetic retinopathy is a severe vision complication of
diabetes and it is one of the main causes of blindness all over the
world. Early detection of diabetic retinopathy is essential to cope
with this adverse effect. Microaneurysm is one of the early signs
of diabetic retinopathy. So the presence of microaneurysm
detection is a prerequisite for early diagnosis of diabetic
retinopathy. In this paper, we have proposed a simple
morphological method for the detection of microaneurysm that
uses top-hat transform. Our method can detect the faint
microaneurysm at low resolution due to contrast enhancement
and noise reduction as preprocessing. We also compare the
results of different retinopathy detection techniques.
Keywords-Diabetic retinopathy, microaneurysm, blindness,
contrast enhancement, morphological operation.
I. INTRODUCTION
Diabetic retinopathy is the most common cause of
blindness. It is one of the consequences of diabetes. Around 7%
people who have diabetes for 10 years will have developed
diabetic retinopathy. The rate of blindness in global population
from diabetic retinopathy will rise to 4.4% at the end of 2030
[1], [2], [3].
Microaneurysm is one of the earliest symptoms of the
diabetic retinopathy. If we can detect the microaneurysm at an
earlier stage then diagnosis of retinopathy can be done
effectively. As a result the treatment of diabetic retinopathy can
reduce the threat of blindness by 50% [4]-[8].
Microaneurysm [4], [9]-[10] is a retina lesion which is
caused by local swelling of capillary walls and generate small
red dots on the surface of the retina shown in Figure 1. It
ranges from 25 – 100 microns in size.
The detection of microaneurysm is difficult due inherent
low contrast characteristic of eye fundus images. So our aim is
to develop a system for the detection of diabetic retinopathy by
detecting microaneurysm, helping the patients as well as the
doctors for early diagnosis of diabetic retinopathy for reducing
blindness of the diabetic patients.
Sopharak et al. [2] and Prakash and K. Sumathi [4]
proposed microaneurysm detection methods using
mathematical morphology. In this paper, we propose a simple
morphological method for the detection of microaneurysm
using top-hat transform along with image preprocessing. We
compare the results of our method with the results of the above
methods.
Besides this, another method was proposed by Pallawala et
al. [9] based on eigenvectors of affinity matrix. We also
compared the results with our results.
Section II describes medical knowledge of diabetic
retinopathy, Section III presents our proposed method, Section
IV shows the experimental results and discussions and finally,
Section V draws the conclusions of our paper.
(a) (b)
(c) (d)
(e)
Figure 1. Abnormal findings in the eye fundus images caused by
diabetic retinopathy: (a) hard exudates, (b) soft exudates, (c)
hemorrhage, (d) microaneurysm and (e) neovascularizations
(images taken from references [11]).
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 8, August 2017
200 https://sites.google.com/site/ijcsis/
ISSN 1947-5500
2. Figure 2. Flow diagram
II. MEDICAL KNOWLEDGE OF DIABETIC RETINOPATHY
Diabetic retinopathy is a consequence of diabetes. It
affects retina’s blood vessels and prevents light entering the
retina that may create blindness [3], [11] and [12] of diabetic
patents. Sometimes these vessels may dilate or close
completely. Figure 1 shows the symptoms of diabetic
retinopathy. In severe retinopathy, the blood vessels may
become diluted which create microinfarcts in the retina. Lack
of oxygen in the microinfarcts cause the development of new
fragile vessels. This phenomenon is called neovascularization
– a serious cause of vision loss.
Microaneurysm is the earliest symptoms of ratinopathy. It
looks like as small, dark red spots and causes intra retinal
haemorrhages. It’s size may vary from 10-100 microns.
III. PROPOSED METHOD
The flow diagram of our method is shown in Figure 2. At
first we have to take a color fundus image as an input. Then the
image is converted into green band image, G. We used median
filtering (to get B image) and histogram equalization (HE
image) as preprocessing for noise reduction and enhancement
purpose. We get shade corrected (SC) image by subtracting B
image from the green band image G through equation (1).
SC=G-B (1)
The top-hat and binarization operations are applied to the
shade corrected (SC) image to get optic disk and blood vessel
which is named as V.
Then the difference image (D) is obtained by equation
(2)
D=G-SC-V (2)
Then on D image we apply recursive region growing
algorithm to get microaneurysm. The output of the proposed
method is shown in Figure 3.
Color fundus
image of retina
Image
acquisition
Green band Image
(G)
Median filtering
(B) and histogram
equalized image
(HE)
Denoising
and contrast
stretching
Finding shade
corrected image
(SC=G-B)
Blood vessel and
optic disk
detection (V)
Removal of blood
vessel
Microaneurysm
detection by
recursive region
i
Analysis
Detection
(a) (b)
(c) (d)
(e) (f)
(g)
Figure 3. Microaneurysm detection: a) input image (b) green
image, G (c) background image after median Filter, B (d ) image
after histogram equalization, HE (e) shade corrected image, SC=
G-B (f) detected blood vessel image, V (g) detected
microaneurysm.
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 8, Augus 2017
201 https://sites.google.com/site/ijcsis/
ISSN 1947-5500
4. [11] Pavle Prentaˇsi´c, “Detection of Diabetic Retinopathy in Fundus
Photographs”, available online at
https://www.fer.unizg.hr/_download/repository/KDI_Prentasic_Pavle.pd
f (accessed on 112.5.17 ).
[12] Seema Garg, and Richard M. Davis, “Diabetic retinopathy screening
update, available online at
http://clinical.diabetesjournals.org/content/27/4/140.full.pdf (accessed on
June 28, 2013).
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 8, Augus 2017
203 https://sites.google.com/site/ijcsis/
ISSN 1947-5500