�ݺ�ߣ

International Journal of Computer Trends and Technology (IJCTT) – volume 7 number 1– Jan 2014
ISSN: 2231-2803 http://www.ijcttjournal.org Page21
Early Proliferation Stage of Detecting Diabetic Retinopathy Using
Bayesian Classifier Based Level Set Segmentation
S.Vijayalakshmi1
, P.Sivaprakasam2
1
(Research Scholar, Karpagam University,Coimbatore, India)
1
(Department of MCA, Park College of Engineering and Technology, Coimbatore, India)
2
(Department of Computer Science, Associate Professor Sri Vasavi College, Erode, India)
ABSTRACT : This article presents Bayesian
Classifier which controls the levels set
segmentation and it detect the retinal clots at an
early stage from the image captured from fundus
camera. The classifier is a probabilistic and used
for the control of level set contour propagation for
the detection of class clot defined, extracting the
retinal vessels even with minute deformation due to
the clots. The algorithm is tested in MATLAB on
fundus images taken at various stages of
progression and results which proves the
effectiveness of the proposed method.
Keywords -Bayesian Classifier, Blood Clots,
Contour Propagation, Diabetic Retina, Level set
segmentation
1. INTRODUCTION
A diabetic retinopathy is the most common
diabetic eye disease which leads to the blindness
because of the changes in the blood vessels of the
retina. In this disease, blood vessels may swell,
leak fluid and abnormal new growth on the surface
of the retina. The retina is the light-sensitive tissue
at the back of the eye. A healthy retina is necessary
for good vision. Figure 1 shows the diabetic retina
with swollen nerves. This makes the vision
impaired and lead to permanent blindness at the
end stages.
Fig 1: A diabetic retinal image taken from a
fundus camera
To detect the diabetic retina at an early
stage it is important to adopt an enhanced image
processing tool for the detection of even small clots
for an early diagnosis. Atul Kumar[2]
identified the
feature of exudates from the image. On basis of
their pixels intensity and frequency it classified into
moderate stage of NPDR. Accuracy to the
extracting feature is then tested with the perception
of the ophthalmologist’s. Firstly raw dataset
(Fundus Retinal Image) is pre-processed by
morphological technique as images are of variant
size, colour contrast and resolution. Then adaptive
threshold and centroid is calculated by Otsu
methodology so the Image boundary is traced.
Then optic disk is localised by calculating ROI
using Hough Transformation and distant from the
image as the intensity of exudate and the optic disk
is same in the fundus image. The SVM classifier
uses features extracted by combined 2DPCA
instead of explicit image features as the input
vector Combined 2DPCA is proposed and then for
acquiring higher accuracy of classification we can
use virtual SVM. N.S.Datta[3]
discussed primarily
the hardware based issues on early detection of
diabetic retinopathy. Software based algorithms for
preprocessing, segmentation and classification
stages are initially analyzed. Later those techniques
were customized and implemented using
TMS320C6713 based DSP Kits of Texas
instruments with code composer studio for the
early detection of diabetic retinopathy through the
fundus images of retina. N.Santhi[4]
described
algorithm to detect the blood vessels effectively by
enhancing the retinal image edges using shearlet
transform. The directionality feature of the multi
structure element makes it an effective tool in edge
detection. Morphological operators using multi
structure elements are applied to the enhanced
image to find the retinal image ridges. By
reconstructing the morphological operators, the
ridges not belonging to the result unchanged. The
remaining ridges belonging to the vessels are
determined by using a thresholding method.

Chunming Li[5]
proposed a novel region-
based method for image Segmentation which is
able to deal with intensity inhomogeneities in the
segmentation. First, based on the model of images
with intensity inhomogeneities, he derived a local
intensity clustering property of the image
intensities, and define a local clustering criterion
function for the image intensities in a neighborhood
of each point. This local clustering criterion
function is then integrated with respect to the
neighborhood center to give a global criterion of
image segmentation. In a level set formulation, this
criterion defines an energy in terms of the level set
functions that represent a partition of the image
domain and a bias field that accounts for the
intensity inhomogeneity of the image. Therefore,
by minimizing this energy, method is able to
simultaneously segment the image and estimate the
bias field, and the estimated bias field can be used
for intensity inhomogeneity correction (or bias
correction). The method has been validated on
synthetic images and real images of various
modalities, with desirable performance in the
presence of intensity inhomogeneities.
Hafiane,A..Bunyak[6]
proposed two forms of
deformable models, the parametric form referred as
snakes, an explicit parametric representation of the
curve is used. This form is compact, robust for
image noise and boundary gaps as it constrains the
extracted boundaries to be smooth. It restricts the
degree of topological adaptability of the model,
especially the deformation involves splitting or
merging of parts. In contrast the implicit
deformable models are designed to handle
topological changes naturally. This form is not
robust to boundary gaps and suffer from several
other deficiencies.
Sau Dan Lee[7]
used naives bayes
classification of uncertain data with a pdf. The key
solution is to extend the class conditional
probability estimation in the Bayes model to handle
pdf’s. Extensive experiments on UCI datasets show
that the accuracy of naïve bayes model can be
improved by taking into account the uncertainty
information. In the proposed method a level set
segmentation used for the retinal image
segmentation. The problem is normally solved by
iterative methods which obtain the current pixel
position using the boundary at the previous time
step. Therefore, it may be required to specify an
initial guess to start the iterative process. A contour
is implicitly represented by the zero level kept as
initial value and evolved according to a motion
equation. To move the contour put on chain rule in
order to derivative inside and outside function.
Level set evolution is used to deduct for finding the
abnormal clots. The Bayesian risk is formed by
false positive and false negative fractions in a
hypothesis test finally it is s used find the iterative
level set function from the derived function.
Therefore results show faster convergence and
efficiency in detected the minute clots.
2. MATERIAL AND METHODS
Image processing techniques as level set
segmentation can be used for the retinal image
segmentation due to the fact that the contours area
propagation and the speed of propagation both are
controllable.
2.1 Methodology
Fig 2 Pre processing and Image
Segmentation
2.2 Level set segmentation
In a level set method the level set surface
has to be moved. In order to move the velocity field
F is defined, shows how contour points move in
time. The movement is based on the application
specific physics such as the time, position, normal,

curvature and most importantly the image gradient
magnitude. The figure3 expresses the method of
the level set algorithm. Consider two conical
surfaces as shown.
Fig 3. The contour propagation method for
the level set segmentation.
In the figure 3 shows the curve is moved up by
setting the level set function (x,y,t) so that it
rises, falls and expands through the blue colour
defined constrained space. After declaring the
dependent parameters it is important to build an
initial value of the level set function
(x,y,t=0) …(1)
which is based on the initial contour position. The
next step is to adjust the over time to time, such
that the contour at time t is defined as
(x(t),y(t),t=0 …(2)
Formulation of the level set
During the formulation of the level function the
importance is given to the constraints involved in
the propagation of the curve. The first constraint is
the level set value of a point on the contour with
motion x(t) must always be zero.
(x(t),t=0) …(3)
Upon application of the chain rule.
+delta(x(t),t).x’(t)=0 … (4)
Since the value of F supplies the speed in the
outward normal direction
X’(t).n=F, where n=delta( ) / | delta( )|
…..(5)
Hence the equation for the evolution of the curve is
+ F| delta( )| =0 ….(6)
2.3 Bayesian classifier application to Level
set segmentation
A Bayes classifier is a simple probabilistic
classifier based on applying Bayes' theorem (from
Bayesian statistics) with strong (naive)
independence assumptions. A more descriptive
term for the underlying probability model would be
"independent feature model”. In simple terms, a
naive Bayes classifier assumes that the presence (or
absence) of a particular feature of a class is
unrelated to the presence (or absence) of any other
feature.
A Bayesian classifier model as per the theorem is
given as
The classifier could be in simple given as
is the probability of the instance d being
in class Cj.
is the probability of generating instance d
given in class Cj.
is the probability of occurrence of class Cj.
p(d) is the probability of instance d occurring.
Assuming that we have two classes C1 and C2
where
C1= normal dimension retinal blood vessels and
C2=abnormal dimension retinal blood vessels
suspected to be a retinal blood clot. The classifier
can be used to propagate the curve in a level set
segmentation algorithm to classify the normal
retinal blood vessels and abnormal retinal clots at
an early stage. The level set function propagation
vector can be unified in the form
+ F| delta( )|
Here the is the contour propagator function
traversing through the retinal pixel image. The
level set function can be enhanced by forming a
likelihood model using a Bayes classifier for
contouring the curves around the retinal clots as
given in equation below.

Assuming contour to be travelling in a direction
with possible next pixel given by
The classifier can hence be given as
3. Results and Discussion
The proposed algorithm is implemented in
MATLAB and the results are analyzed for cases
like end stage proliferation and beginning of the
proliferation stage. Figure 4 shows the original
input image converted to gray scale image .The
application of a Bayesian classifier aided in
detection of the blood vessels with a cluster of
blood clots. The cots with clustering of 0.15 mm
were successfully detected . Some of the clots
above the cluster with 0.03 mm dimensions were
also detected suggesting the effectiveness of the
algorithm for the macular clot detection.
Fig 4: Extracted blood vessels for the end
proliferative stage
Figure 5 shows the early stage proliferation of the
diabetic retina. The algorithm showed the
efficiency in detected the clots of size 0.02mm near
the optic disc which was about to proliferate to the
next stage. Clots with minute dimensionality at the
micro macular vessels were detected at two places.
Fig 5: Extracted blood vessels for the start of
proliferative stage
4. Conclusion
The accurate extraction of the retinal
blood clots can provide much help for diagnosis of
diabetic retinopathy diseases. Even though many
techniques and algorithms have been developed,
there is still room for improvement in finding the
minute blood clots in segmentation methodologies.
In this paper, we have presented a Bayesian
classifier to detection of the blood vessels with a
cluster of blood clots. The proposed method has
been tested on real retinal image databases and the
experimental results have been compared with
those of other methods. Since the developed
algorithm detected in micro level clots
information into account, it can result in very good
performance in the detection of early stage of
proliferation of the diabetic retinal eye. The result
demonstrates that the proposed method
outperforms other methods by deducting the
minute clots in early proliferation stage.
References
[1] http://www.nei.nih.gov/health/diabetic/retinopathy.asp
[2] Atulkumar, Manish Srivastava, A.K.Sinha, A segment based
technique for detecting exudate from retinal fundus image,
International Journal of Computer Science & Engineering
Technology, Vol 3 No.7,2012,pp.274-284.
[3] N.S.Datta, R.Banerjee, H.S.Dutta,S.Mukhopadhyay,
Hardware based analysis on automated early detection of
diabetic retinopathy, Publised by Elsevier Ltd, Procedia
Technology,2012,pp.256-260.
[4] N.Santhi, G.D.Shanjula, K.Ramar, A Novel Algorithm To
Analyze Retinal Image Using Morphological Operators,
International Conference On Modelling Optimizations and
Computing,2012,pp.3880-3891.

[5] Chunming Li, Rui Huang, Zhaohua Ding, J.Chris Gatenby,
Dimitris N. Metaxas,John C.Gore, A Level Set Method for
Image Segmentation in the Presence of Intensity
Inhomogeneities with Application to MRI, IEEE Transactions on
Image Processing,Vol 2,2011,pp.2007-2016.
[6] Hafiane, A, Bunyak, F. ; Palaniappan, K. Evaluation of
level set-based histology image segmentation using geometric
region criteria, Biomedical Imaging: From Nano to Macro,
2009. ISBI '09. IEEE International Symposiumpp,2009,1-4.
[7] Jiangtao Ren , Sau Dan Lee ; Xianlu Chen ; Ben
Kao ; Cheng, R. ; Cheung, D.Naives Bayes classification of
uncertain data. Data Mining, 2009. ICDM '09. Ninth IEEE
International Conference.2009,pp.944-949.
[8] L. F. Costa and R. M. Cesar-Jr, Shape Analysis and
Classification:Theory and Practice. Boca Raton, FL: CRC,
2001.
[9] Tenn Francis Chen Medical Image Segmentation using
Level SetsTechnical Report #CS-2008-12.

�ݺ�ߣ

Ijctt v7 p104

Recommended

More Related Content

What's hot (17)

Similar to Ijctt v7 p104 (20)

Ijctt v7 p104