International Journal of EmergingTrends & Technology in Computer Science(IJETTCS)

Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com

Volume 3, Issue 2, March April 2014 ISSN 2278-6856


Volume 3, Issue 2 March April 2014 Page 1


Abstract: The f i el d of medi cal i magi ng gai ns i t s
i mport ance wi t h i ncrease i n t he need of accurat e
and ef f i ci ent di agnosi s over a short peri od of
t i me. MR i magi ng has become a wi del y - used
met hod of hi gh qual i t y medi cal i magi ng,
especi al l y i n brai n i magi ng where MR' s sof t
t i ssue cont rast and non- i nvasi veness are cl ear
advant ages. MRI segment at i on i s an i mport ant
i mage processi ng st ep t o i dent i f y anat omi cal
areas of i nt erest f or di agnosi s of many di sorders
such as brai n t umor, mul t i pl e scl erosi s, et c.
Segment at i on approaches have met wi t h onl y
l i mi t ed success because of overl appi ng i nt ensi t y
di st ri but i ons of i nt racrani al and ext racrani al
t i ssues i n order t o make robust aut omat i c brai n
t umor and heal t hy t i ssue segment at i on. Hence t he
t he ext racrani al t i ssues shoul d be removed i n
brai n MR i mages and maki ng use of onl y
i nt racrani al t i ssue regi ons f or segment at i on of
t umor and normal t i ssues f or f urt her anal ysi s. In
t hi s paper, f rom t he abnormal MR i mages, t he
f eat ures are ext ract ed. Here, t he f eat ure
ext ract i on i ncl udes t he f i rst order and second
order f eat ures. Fi rst order t ext ure measures are
st at i st i cs cal cul at ed f rom t he ori gi nal i mage
val ues, l i ke vari ance, and do not consi der pi xel
nei ghbour rel at i onshi ps. Second order measures
consi der t he rel at i onshi p bet ween groups of t wo
( usual l y nei ghbouri ng) pi xel s i n t he ori gi nal
i mage. The pri nci pal component s are sel ect ed.
Then, an ef f i ci ent segment at i on al gori t hm f or
magnet i c resonance i mages of brai n t i ssues usi ng
f uzzy l ogi c i s proposed. The f uzzy l ogi c out put
wi l l be compared wi t h K- means, Neural Net work
f or segment at i on. The comparat i ve anal ysi s wi l l
be done i n t erms of perf ormance measured
paramet ers.

Keywords: Image segmentation, k mean, fuzzy logic,
clustering, neural network.

1. INTRODUCTION
Magnetic resonance imaging (MRI) is used as a valuable
tool in the clinical and surgical environment because of
its characteristics like superior soft tissue differentiation,
high spatial resolution and contrast and it does not use
harmful ionizing radiation to patients. The data obtained
from MR images are used for detecting tissue deformities
such as tumors, cancers and injuries. In order to
understand MRI contrast, it is important to have some
understanding of the time constant involved in relaxation
processes that establish equilibrium following RF
excitation.Time constants involved in MR images are
T1(realign time), T2(relaxation time) and T2f(flair). The
tumour detected using MRI are the following Primary
malignant tumour (glioma and meningioma) and
secondary malignant tumours (metastatic brain tumor).
MRI segmentation is an important image processing step
to identify anatomical areas of interest for diagnosis of
many disorders such as brain tumour, multiple sclerosis,
etc.
1.1. Feature extraction: To classify an object in an
image, we must first extract some features out of the
image. Feature extraction is a special form of
dimensionality reduction and features reflect properties
measured at the pixel-level that can aid in discriminating
between normal pixels and tumor pixels. First order
texture measures are statistics calculated from the original
image values, like variance, and do not consider pixel
neighbour relationships. Second order measures consider
the relationship between groups of two (usually
neighbouring) pixels in the original image.
1.2. Cluster analysis: Cluster analysis is an
interdependence technique. It is similar to multi
dimensional scaling .The difference is that multi
dimensional scaling identifies underlying dimensions,
while cluster analysis identifies clusters. The goal of
clustering is to reduce the amount of data by categorizing
or grouping similar data items together. Clustering
approaches have met with only limited success because of
overlapping intensity distributions of intracranial and
extra cranial tissues and hence the extracranial tissues are
to be removed.

2. Overview of other method
Clustering can be considered the most important
unsupervised learning problem, so it deals with finding a
structure in a collection of unlabeled data. A cluster is
therefore a collection of objects which are similar
between them and are dissimilar to the objects
belonging to other clusters. Clustering algorithms may be
classified as listed below
2.1. K-Means segmentation
K-means is one of the simplest unsupervised learning
algorithms that solve the well known clustering problem.
The main advantages of this algorithm are its simplicity
and speed, which allows it to run on large datasets. The
procedure follows a simple and easy way to classify a
given data set through a certain number of clusters fixed a
Segmentation Of Medical Image Using
Fuzzy Neuro Logic

P.Priyadharsini
1
, I .Rexiline Sheeba
2

1
PG scholar, Sathyabama University, Chennai, India.
2
Assistant Professor /ECE Department, Sathyabama University, Chennai, India.

International Journal of EmergingTrends & Technology in Computer Science(IJETTCS)
Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 3, Issue 2, March April 2014 ISSN 2278-6856


Volume 3, Issue 2 March April 2014 Page 2


priori. The main idea is to define k centroids , one for
each cluster. These centroids should be placed in a
cunning way because of different location causes different
result. So, the better choice is to place them as much as
possible far away from each other. The next step is to take
each point belonging to a given data set and associate it to
the nearest centroid. When no point is pending, the first
step is completed and an early group age is done. At this
point we need to re-calculate k new centroids as bary
centers of the clusters resulting from the previous step.
After we have these k new centroids, a new binding has to
be done between the same data set points and the nearest
new centroid. A loop has been generated. As a result of
this loop we may notice that the k centroids change their
location step by step until no more changes are done. In
other words centroids do not move any more.
Disadvantage of this algorithm is that the resulting
clusters depend on the initial assignments. But it does
not ensure that the solution given is not a local minimum
of variance. Several misclassified data points after
segmentation of brain image. The K means algorithm is
given below
Step1: Choose K Initial Centers Z
1
(1), Z
2
(2)--- They are
arbitrary
Step2: At the K
th
iterative Step, distribute the sample {X}
among the K Cluster domain
,using the relation X S
j
(k)
if || X- z
j
(k) || < || X-z
i
(k) ||, Where s
j
(k)-the
set of samples whose cluster center is z
j
(k)
Step3: from the result of step 2 , calculate the new
clusters z
j
(k+1),j=1,2-----kz
j
(k+1)=1/n
j
x X
S
j
(k)where n
j
number of samples in s
j
(k),cluster centers
are sequentially updated.
Step4: if z
j
(k+1)= z
j
(k) ,the algorithm has converged
and procedure is terminated otherwise go to step 2.

3. PROPOSED METHOD
3.1. Fuzzy Logic
Fuzzy set theory provides a host of attractive aggregation
connectives for integrating membership values
representing uncertain information. These connectives can
be categorized into the following three classes union,
intersection and compensation connectives. The
membership function of a fuzzy set in a functional form,
typically a bell-shaped function, triangle-shaped function,
trapezoid-shaped function, etc.
When fuzzy systems are applied to appropriate problems,
particularly the type of problems described previously,
their typical characteristics are faster and smoother
response than with conventional systems. This translates to
to efficient and more comfortable operations for such tasks
tasks as controlling temperature, cruising speed, for
example. Furthermore, this will save energy, reduce
maintenance costs, and prolong machine life. In fuzzy
systems, describing the control rules is usually simpler and
and easier, often requiring fewer rules, and thus the
systems execute faster than conventional systems. Fuzzy
systems often achieve tractability, robustness, and overall
low cost.
The procedure for obtaining the fuzzy output of such a
knowledge base can be formulated as 1. The firing level of
of the i-th rule is determined by
Ai(x0) Bi(y0).
2. The output of of the i-th rule is calculated by
C_i(w) :=Ai(x0) Bi(y0) Ci(w)
for all w W.
3. The overall system output, C, is obtained from the
individual rule
Outputs C_i by
C(w) =Agg{C_1, . . . , C_n}for all w W.

The methodology proposed in this paper is as explained
in Fig:

Fig 1 Flow Diagram for the proposed methodology

Here, in this paper, basically, the clustering of abnormal
brain MR images is done using fuzzy logic which is the
proposed method. From the abnormal brain MR images
the first order and the second order features were
extracted. First Order Features include mean, variance,
skewness , kurtosis, energy and entropy while the second
order features include angular second momentum,
entropy, contrast, cluster shade, cluster prominence,
inertia, and local homogeneity. The features were
extracted and using them a feature vector was formed.
Then the selected features are segmented using Fuzzy
logic. Then, a comparative study is made using K-means,
neural network and Fuzzy logic.

4. Result and Conclusion:
The various clustering methods are analyzed. These
methods are used to perform tissue classification in MRI.
The experimental results are shown in figure 4.1. The
major disadvantage of K means algorithm had several
misclassified data points after segmentation of brain
image.

International Journal of EmergingTrends & Technology in Computer Science(IJETTCS)
Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 3, Issue 2, March April 2014 ISSN 2278-6856


Volume 3, Issue 2 March April 2014 Page 3



Figure 2 input image


Figure 3 Clustering Using K-MEANS
(No of clusters=3)


Figure 4 Clustering Using FUZZY
(No of clusters=3)

The following table shows the misclassification rate
performed by K-means and Fuzzy Logic methods

Table 1:Misclassification rate
Method MR
White
MR
Gray
MR
CSF
K-Means 30 35 99

MR misclassification rate

References
[1] J.G.Webster, Ed. , Medical Instrumentation:
Application and Design.New York:John Wiley &
Sons, Inc. , 1998, pp.551-555.
[2] Haack.E et al. , 1999, Magnetic Resonance
Imaging, Physical Principles and Sequence Design.
Wieley-Liss, New York
[3] Rui Xu, and Donald Wunsch II,Survey of Clustering
Algorithms,IEEE Transaction on Neural Networks,
VOL.16, NO. 3, MAY 2005
[4] Boris Cigale and Damjan Zazula, Segmentation of
ovarian ultrasound images using cellular neural
networks, International Journal of Pattern
Recognition and Artificial Intelligence Vol.18
[5] A.Slavova and V. Mladenov, Cellular Neural
Networks: Theory and Applications, Nova Science
Publishers, Inc., USA, 2004
[6] X.Zhang,X.L.Xiao,J .W.Tian,J .Liu and G.Y.Xu,
Application of support vector machines in
classification of Magnetic Resonance
Images,International journal of computers and
applications,vol 28,No.2.2006.
[7] Mark Schmidt, Automatic Brain Tumor
Segmentation, Master Thesis,University of
Alberta,CANADA,2005
[8] Boris Cigale and Damjan Zazula, Segmentation of
ovarian ultrasound images using cellular neural
networks, International Journal of Pattern
Recognition and Artificial Intelligence Vol.18
[9] M. Hanggi and G. S. Moschytz, Kluwer,Cellular
Neural Networks, 2000, International Journal of
Electrical Engineering Education.