Mask image generation for segmenting retinal fundus image features into isnt quadrants using array centroid method

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
__________________________________________________________________________________________
Volume: 03 Issue: 04 | Apr-2014, Available @ http://www.ijret.org 263
MASK IMAGE GENERATION FOR SEGMENTING RETINAL FUNDUS
IMAGE FEATURES INTO ISNT QUADRANTS USING ARRAY
CENTROID METHOD
Darsana S1
, Rahul M Nair2
1
MTech Student, Department of ECE, Nehru College of Engineering and Research Centre Thrissur, Kerala, India
2
Assistant Professor, Department of ECE, Nehru College of Engineering and Research Centre Thrissur, Kerala, India
Abstract
Retinal fundus image processing paves the way for diagnosis of many eye disorders. The fundus image features such as optic disc,
optic cup, optic rim and blood vessels help to analyse the pathological condition of eye. The analysis involves calculation of ocular
parameters such as Rim to Disc Ratio (RDR) and Inferior Superior Nasal Temporal (ISNT) Ratio and verification of ISNT Rule. For
this analysis the fundus image features have to be segmented into ISNT quadrants. This paper presents a novel approach towards
mask image generation for segmenting fundus image feature into ISNT quadrants using array centroid method. The existing method
uses a predefined mask image which is not accurate and the proposed array centroid method overcomes the limitations of existing
method. The method involves centroid calculation, array initialization; mask image generation and, mask- feature image
multiplication. The proposed method is applied to the fundus images collected from a reputed eye hospital.
Keywords: Fundus Image Features, Array Centroid Method, Centroid Calculation, Mask Image Generation
-----------------------------------------------------------------------***-----------------------------------------------------------------------
1. INTRODUCTION
Image processing is a vital field of research and has been
established as a core field of innovation in modern health care.
Among the processing techniques medical image
segmentation plays an important role in biomedical
applications. The retinal fundus image processing helps to
identify eye disorders such as glaucoma, diabetc retinopathy,
hypertension, macular edema, arteriosclerosis etc. Fundus
image processing involves fundus image feature extraction
and its analysis. Inorder to analyse the fundus features some
ocular parameter have to be calculated. For this the features
have to be segmented into the ISNT quadrants. In the
existing methods predefined mask images of size same as that
of the image are used to segment the features into four
quadrants. In such cases centroid of the feature is not
considered and thus results in inaccurate segmentation of the
features. This paper proposes a novel approach towards mask
generation for ISNT quadrants using array centoid method.
This method involves optic disc centroid calculation, array
initialization, mask generation and mask-feature image
multiplication.
The paper is organized as follows. The section 2 gives
theoretical background which includes basics of fundus eye
image, fundus image features and ISNT quadrants. Section 3
describes the proposed array centroid method. Section 4 shows
the experimental results. Conclusion and future scope are
presented in section 5.
2. THEORETICAL BACKGROUND
The basics of fundus image, fundus image features and ISNT
quadrants are explained below.
2.1 Fundus Image
The fundus of the eye is the interior surface of the eye,
opposite to the lens, and includes the retina, optic disc,
macula, fovea, and posterior pole. Fundus photography (also
called fundography) is the creation of a photograph of the
fundus. Modern fundus photographs generally recreate
considerably larger areas of the fundus than what can be seen
with handheld ophthalmoscopes. Fundus photography is
performed by a fundus camera, which basically consists of a
specialized low power microscope with an attached camera.
Fundus image for a left eye is shown in Fig 1.
Fig -1: Fundus eye image

__________________________________________________________________________________________
2.2 Fundus Image Features
The important fundus image features are optic disc, optic cup,
optic rim and blood vessels as shown in Fig2. The optic disc
or optic nerve head is the location where ganglion cell axons
exit the eye to form the optic nerve. The optic disc is also the
entry point for the major blood vessels that supply the retina.
There is a central depression, of variable size, called the optic
cup.
Fig -2: Fundus eye image features
Optic rim is the region between the optic disc and optic cup.
The pattern of rim width in a healthy optic nerve is thickest
inferiorly, followed by superiorly, nasally and the temporal
rim is the thinnest.
2.3 Extracted Fundus Eye Features
The fundus eye image features can be extracted from the
fundus color image using image processing techniques such as
color space analysis, morphological processing, filtering and
thresholding. The binary images of fundus features for the
image shown in Fig 1 is as shown in Fig 3
(a) (b) (c)
Fig -3: Binary images of fundus image features (a) Optic disc
(b) Optic rim (c) Blood vessels in disc
2.4 Fundus Image
The Inferior Superior Nasal and Temporal quadrants for
retinal image are as shown in Fig 4. Inferior quadrant is the
lowermost quadrant and superior quadrant is the uppermost
quadrant. The nasal quadrant is the right quadrant in right eye
and left quadrant in left eye. The temporal quadrant is the left
quadrant in right eye and right quadrant in left eye.
Fig -4: ISNT quadrants
The relevance of these quadrants in fundus image analysis is
that the segmented fundus eye features in the ISNT quadrants
can be used for the calculation of ISNT ratio, Rim to Disc
Ratio (RDR) and ISNT rule verification which further helps in
the diagnosis of many eye disorders. The cardinal clinical sign
of glaucomatous optic neuropathy is thinning of the
neuroretinal rim, with the regional preference for the superior
and inferior poles of the optic disc. RDR is calculated for the
inferior and superior quadrant as the ratio of rim area to disc
area. ISNT ratio is calculated using the ratio of area of blood
vessels in inferior superior to nasal temporal side of optic disc.
There may be a shift in blood vessels to nasal side in
glaucoma patients and thus the ISNT ratio will be less
compared to normal person’s image. The ISNT rule states that
in normal eyes the thickness of the neuroretinal rim along the
cardinal meridians of the optic disc ,that is the rim width,
decreases in the order inferior (I), superior (S), nasal (N),
temporal (T), and that the neuroretinal rim in glaucomatous
optic disc violates this quantitative relationship. The four
variants of ISNT rule are (I>S, I>S>T, I>T, and combined I>T
and S>T).
3. METHODOLOGY
The flow chart for the array centroid method is shown in Fig
5. The method involves centroid calculation, array
initialization, mask image generation and mask- feature image
multiplication.

__________________________________________________________________________________________
Fig -5: Flow chart of Array Centroid Method
3.1 Centroid Calculation
The first step in array centroid method is the calculation of
centroid of the optic disc. The centrod value of optic disc is
calculated and the centroid is located in the image as shown in
Fig 6. The centroid values are denoted as (C1, C2).
Fig -6: Centroid of optic disc
3.2 Array Initialization
The Array consists of rows and columns and the size of the
array is defined by this rows and columns. The array
initialization involves defining an array of size same as that of
the image to be segmented and then initializing all its values
with zero as shown in Fig 7. The image size is defined by its
pixels and each pixels are denoted as (i,j).The number of rows
and columns in array are denoted as R and C respectively.
Fig -7: Initialized array
3.3 Mask Image Generation
The mask image for each quadrant is generated exploiting the
characteristics of array and the centroid value. The mask
image act as an image filter Thus the mask image should be
such that the desired quadrant should be white and all other
quadrants should be black. The mask generation for the four
different quadrants are explained below.
Inferior quadrant is the lowermost quadrant. The mask for this
quadrant should be such that inferior quadrant is white or
binary 1 and all other quadrants are black or binary 0.The
logic used to generate inferior quadrant mask is given below.
Superior quadrant is the uppermost quadrant and the mask
image should be such that superior quadrant is white or binary
1 and all other quadrants are black or binary 0. The logic used
to generate superior quadrant mask is given below.
The leftmost quadrant is the nasal quadrant for left eye and
temporal quadrant for the right eye. The mask for left
quadrant has white pixels on left quadrant and black pixels on
other quadrants. The logic used for the mask generation of left
quadrant is shown below.
The rightmost quadrant is the temporal quadrant for left eye
and nasal quadrant for the right eye. The mask for right
quadrant has white pixels on right quadrant and black pixels
on other quadrants. The logic used for the mask generation of
right quadrant is shown below.
For i varies from C2 to R and j varies from 1 to C,
set the pixels 0 for following condition and all other
pixels as 1
C1=C2 i+j<C1+C2 or i<j
C1>C2 i+j<C1+C2 or i<j and j-1>C1-C2
C1<C2 i+j<C1+C2 or i<j or i=j or i-j<C2-C1
For i varies from 1to C2 j varies from 1 to C, set the
pixels 0 for following condition and all other pixels as 1
C1=C2 i+j>C1+C2 or i>j
C1>C2 i+j>C1+C2 or i>j or i=j or j-1<C1-C2
C1<C2 i+j>C1+C2 or i>j and i-j>C2-C1
For i varies from 1to R j varies from 1 to C1,set the
C1=C2 i+j>C1+C2 or i<j
C1>C2 i+j>C1+C2 or i<j and j-1>C1-C2
C1<C2 i+j>C1+C2 or i<j or i=j or i-j<C2-C1

__________________________________________________________________________________________
3.4 Mask Feature Image Multiplication
Inorder to segment the fundus features into the ISNT
quadrants the feature image have to be multiplied with the
respective quadrant mask image. Thus the optic disc , optic
rim and blood vessel in each of the four quadrants are
obtained.
4. EXPERIMENTAL RESULTS
The proposed method is applied to the image database
collected from hospital. The different fundus image features
are segmented into the four quadrants using array centroid
method.
The mask generated for the ISNT quadrants are shown in Fig
8
(a) (b) (c) (d)
Fig -8: Mask generated for (a) inferior quadrant (b) superior
quadrant (c) left quadrant (d) right quadrant
The retinal fundus image features such as optic disc , optic rim
and blood vessels in disc region are segmented into the ISNT
quadrants using the mask generated and the segmented results
are shown in Fig 9 .
Fig 9: Segmented optic disc optic rim and blood vessels in
ISNT quadrants
5. CONCLUSIONS
In this paper, a novel approach for the mask image generation
for segmenting retinal fundus image features into ISNT
quadrants has been presented. The proposed array centroid
method have four steps, centroid calculation, array
initialization, mask image generation and mask-feature image
multiplication. The centroid of optic disc is calculated and
then an array of size same as that of the image is initialized
with all its elements zero. Then the masks are generated for all
the four quadrants using the array characteristics and centroid
value. The generated mask is multiplied with the feature
image to obtain the quadrant segments. The proposed method
accurately segments the optic disc, optic rim and blood vessels
in disc region into ISNT quadrants.
The segmented features can be used for the ocular parameter
calculation. The ocular parameters that can be calculated from
the obtained features are Cup to Disc Ratio (CDR), Rim to
Disc Ratio(RDR), Inferior, Superior Nasal, Temporal (ISNT)
ratio etc.Further these parameters help in the diagnosis of eye
disorders like glaucoma.
ACKNOWLEDGEMENTS
The authors are grateful to Dr. Rajesh Radhakrishnan. M. S
Ophtalmic Surgeon and Glaucoma Consultant, Adithya Kiran
Eye Care Centre Palakkad for providing the fundus image
photographs and the guidance and support given for this work.
REFERENCES
[1]. C. B. Anusorn, W. Kongprawechnon, T. Kondo, S.
Sintuwong and K. Tungpimolrut, “Image Processing
Techniques for Glaucoma Detection Using the Cup to Disc
Ratio,” Thammasat International Journal of Science and
Technology. vol.18, no.1, Jan-March 2013.
[2]. S. Kavitha and K. Duraiswamy “An Efficient Decision
Support System For Detection of Glaucoma in Fundus Images
Using ANFIS”, International Journal of Advances in
Engineering & Technology vol.2, Issue.1, pp.227- 240, Jan.
2012
[3]. K. Narasimhan, Dr. K. Vijayarekha, “An Efficient
Automated System For Glaucoma Detection Detection Using
Fundus Image,” Journal of Theoretical and Applied
Information Technology vol.33, no.1 Nov. 2011.
[4]. K. Narasimhan , K.Vijayarekha,K. A. Jogi Narayana,P. S
SivaPrasad and V. SathishKumar, “Glaucoma Detection m
From Fundus Image Using Opencv,” Research Journ al al
of Applied Sciences, Engineering andTechnology, pp.
5459-5463, Dec. 2012.
[5]. S.Kavitha, S.Karthikeyan, Dr.K.Duraiswamy,
“Neuroretinal rim quantification in Fundus images to Detect
Glaucoma,” IJCSNS International Journal of Computer
Science and Network Security., vol.10,no.6, June 2010
[6]. J. Acharya, S.Gadhiya, K. Raviya, “Segmentation
Techniques for Image Analysis: A Review,” International
For i varies from 1toR j varies from C1 to C, set the
C1=C2 i+j<C1+C2 or i>j
C1>C2 i+j<C1+C2 or i>j or i=j or j-1<C1-C2
C1<C2 i+j<C1+C2 or i>j and i-j>C2-C1

__________________________________________________________________________________________
Journal of Computer Science and Management Research.
vol.2, Issue 1, pp.1218- 1221, Jan. 2013
[7]. P. K. Suryawanshi “An Approach to Glaucoma Using
Image Segmentation Techniques” International Journal of
Engineering Sciences & Research Technology, pp. 2388-2392,
Sep 2013.
[8]. J. Kaur, Dr.H.P.Sinha, “An Efficient Blood Vessel
Detection Algorithm For Retinal Images Using Local Entropy
Thresholding,” International Journal of Engineering
Research & Technology, vol.1, Issue.4, pp.1-6 , June. 2012.
[9]. S. Morales, V. Naranjo, J. Angulo and M. Alcaniz,
“Automatic Detection of Optic Disc Based on PCA and
Mathematical Morphology.” IEEE Transactions on Medical
Imaging, vol.32, no.4, April 2013.
BIOGRAPHIES
Darsana S received her BTech degree in
ECE (first class with honours) from Nehru
College of Engineering and Research Centre,
Pampady, Thrissur, Kerala under University
of Calicut in 2012. Presently she is doing
MTech in Applied Electronics and
Communication Systems in Nehru College of Engineering and
Research Centre, Pampady, Thrissur, Kerala.
Rahul M Nair received his BTech degree in
ECE from Ilahia College of Engineering and
Technology, Muvattupuzha, Kerala under
MG University in 2009 and MTech degree in
VLSI and Embedded Systems from VJCET,
Vazhakulam, Kerala under MG University in
2011. Presently he is working as Assistant Professor in Nehru
College of Engineering and Research Centre Pampady
Thrissur, Kerala.

Mask image generation for segmenting retinal fundus image features into isnt quadrants using array centroid method

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Mask image generation for segmenting retinal fundus image features into isnt quadrants using array centroid method (20)

More from eSAT Publishing House (20)

Recently uploaded (20)

Mask image generation for segmenting retinal fundus image features into isnt quadrants using array centroid method