International Journal of Advanced Engineering Research and Technology (IJAERT) 48

Volume 4 Issue 3, March 2016, ISSN No.: 2348 8190

DISCRIMINATIVE CLUSTERING AND FEATURE SELECTION FOR

WML IN MRI BRAIN USING NEURO FUZZY INFERENCE
1

M.Vignesh, 2D.Gayathri, 3M.Kalaiarasi, 4P.Sathya

1
Assistant professor, 2, 3, 4UG Scholar
1, 2, 3, 4
Pollachi Institute of Engineering and Technology

ABSTRACT
White Matter Lesions (WMLs) are minor part of dead
cells found in the parts of the brain. In general, it is
difficult for experts medical to accurately quantify the
WMLs due to decreased contrast between White
Matter (WM) and Grey Matter (GM) in MRI brain
images. The main intention is to detect the White
Matter Lesions present in MRI brain images which
may result in memory loss, stroke or even death.
WML detection process includes the following stages:
1.Image preprocessing, 2.Clustering (Fuzzy c-means
clustering
(FCM),
Geostatistical
Possibilistic
clustering (GPC), Geostatistical Fuzzy clustering
(GFCM) and Neuro-Fuzzy Inference system
(NFIS)).Geostatistical Fuzzy C-means Clustering
(GFCM) algorithm which is 90.17% accurate and less
sensitive to noise over FCM and GPC but there will be
slight noise present at output and it find false lesions
also. To overthrown this and to make detection more
accurate Neuro-Fuzzy inference system is proposed
which is found to be 98.60% accurate over GFCM.
KEYWORDS: Fuzzy clustering, Geostatistics, Image
segmentation,
Magnetic
resonance
imaging,
Possibilistic clustering, White Matter Lesions.

1. INTRODUCTION
Medical imaging is the technique used to
produce images of the human body for clinical or
medical science that produce images of the internal
aspect of the body. MRI of brain is highly sensitive
for detecting all forms of White Matter abnormalities.
Non-specific variation in the White Matter appear
frequently on MRI in elderly patients presenting with
either stroke or cognitive impairment. In regular,
human brain consists of main components namely,
White Matter (WM) and Grey Matter (GM).Neuronal
tissue include mainly long, myelinated axons is known
as White Matter. Myelin is responsible for the white
presence of White Matter. White Matter Lesions
(WMLs) are commonly found in patients with
Multiple Sclerosis (MS), Cerebrovascular Disease
(CVD), stroke, and other neurological disorders.
1.1. WHITE MATTER LESIONS
White matter increases the speed of
transmission of all nerve signals. Its main function is
to regulate electrical signal in axons. White matter is
made of axons and connecting different parts of gray

matter to each point.

White Matter Lesions are minor part of dead
cells found in parts of the brain that act as connectors
are detected using magnetic resonance imaging
(MRI).Evaluation of WMLs in MRI is conventionally
performed using skill and knowledge of experts. This
manual assessment on WML outcome in different
ratings which make it non reproducible and difficult
for a general agreement. Manual evaluation of WM
lesions is not only time consuming but also shows
high inconsistency among human raters. The main
form of remedy is to combat symptoms and slow the
condition's progress using prescription medications.
White matter is actually overwhelmed by a number of
different diseases and conditions, such as stroke,
tumors, lupus, multiple sclerosis and Alzheimers
disease. Controlling blood pressure is an important
way to help stop white matter problems from getting
worse.

2. LITERATURE SURVEY
Anbeek introduced k-nearest neighbors algorithm
(k-NN) for automatic segmentation of WMLs. This is
a supervised learning process and used the information
from T1-weighted, inversion recovery (IR), proton
density-weighted (PD), T2-weighted, and fluid
attenuation IR (FLAIR) scans in order to evaluate the
probability of voxels. T1-weighted scans are a
standard primary scan, in particular differentiating fat
from water - with water darker and fat brighter. This is
one of the basic category of MR contrast and is a
commonly run clinical scan. T2 - weighted scans are
addition basic type. Like the T1-weighted scan, fat is
separated from water - but in this case fat shows
darker, and water lighter.A probability mixture model
and the Bayesian classifier was used by Khayat in
order to extract normal tissue, irregular tissue and
cerebrospinal fluid (CSF) which serves primary needs
like buoyancy, protection and chemical stability.
Normal tissue specifies White Matter and Grey Matter
of brain whereas abnormal tissue refers to lesions of
brain in FLAIR-MR images. This method does not
target on the lesions of small size or irregular shape.
M. L. Seghier, A.Ramlackhansingh presented
a new strategy to identify any type of brain damage
given a single anatomical image. This procedure is
based on the assumption that the lesion comprises a
ordinary voxels that disclose themselves as outliers in
grey and white matter segments.

www.ijaert.org

International Journal of Advanced Engineering Research and Technology (IJAERT) 49

Volume 4 Issue 3, March 2016, ISSN No.: 2348 8190

A ordinary voxels are those that do not

correspond to the expected tissue types; i.e., are no
more Grey Matter (GM), White Matter (WM) neither
Cerebrospinal Fluid (CSF). To avoid misclassification,
they proposed a modified rendition of the unified
segmentation scheme to segment healthful
and
damaged brain tissue.

3. EXISTING METHOD
3.1 FUZZY C-MEANS CLUSTERING
Fuzzy c- means (FCM) is a method of
clustering which grant one piece of data to belong to
two or more clusters. This method (developed by
Dunn in 1973 and improved and revised by Bezdek in
1981) is frequently used in pattern recognition. Fuzzy
c-means has been a very essential tool for image
processing in clustering objects in an image. Fuzzy
c-means (FCM) clustering is an unsupervised
method derived from fuzzy logic that is suitable for
solving multiclass and ambiguous clustering issues.
Fuzzy c-means (FCM) clustering is an unsupervised
technique that has been successfully applied to feature
analysis, clustering, and classifier designs in fields
such as astronomy, geology, medical imaging, target
recognition, and image segmentation.
An image can be described in various
feature spaces, and the FCM algorithm classifies the
image by grouping related data points in the feature
space into clusters. This clustering is achieved by
iteratively decrease a cost function that is
dependent on the distance of the pixels to the
cluster centers in the feature domain.A major
demerit of its use in imaging applications,
however, is that FCM does not incorporate
information around spatial context, causing it to be
sensitive to noise and more imaging artifacts.FCM
provide more false positives in brain image and they
are fewer sensitive to noise. The main one is that the
membership functions are not decreasing with
respect to the distance to the class center.Main
objective of fuzzy c-means algorithm is to minimize
the objective function.
n c
J (U, V) = (ij)m ||xi-vj||2
i=1 j=1
FCM ALGORITHM
Step 1: Randomly select c cluster centers.
Step 2: Compute the Euclidean distance.
Step 3: Calculate the fuzzy membership.
Step 4: Calculate the fuzzy center.
Step 5: Repeat steps (2) and (3) until the minimum J
value.
One disadvantage of FCM is that it computes
the neighborhood term in each iteration step, which is
very time-consuming.

3.2
GEOSTATISTICAL
POSSIBILISTIC
CLUSTERING (GPC)
Although FCM is a very convenient clustering
method, its memberships do not always correspond
well to the degree of belonging of the data and may
be unreliable in a noisy environment. To improve
this weakness of FCM and to produce memberships
that have a good description for the degree of
belonging for the data, Possibilistic approach was
proposed. Possibility clustering algorithms justify
the fact that it can be applied for one cluster at a
time.
GPC ALGORITHM
Step 1: Randomly select c cluster centers.
Step 2: Calculate the possibility membership.
Step 3: Calculate the spatial variability.
Step 4: Incorporate spatial variability into objective
functions.
Step 5: Minimize the objective functions (a
small value of difference) then stop.
3.3 GFCM
The fuzzy C-means objective function is
derived to include a spatial penalty on the
membership functions. The fuzzy C-means algorithm
(FCM) has been utilized in a broad variety of image
processing applications such as medical imaging and
remote sensing.
Its merits include a straight forward
implementation, fairly robust behavior, applicability
to multichannel data, and the capacity to model
uncertainty within the data.
A major demerits of its use in imaging
applications is that FCM not incorporate knowledge
about spatial context, causing it to be sensitive to
noise and other does imaging artifacts. Therefore
Geostatistical Fuzzy C-means clustering is proposed.
The advantages of the new method are the
following:
(1) It yields regions more homogeneous than
those of other methods.
(2) It removes noisy spots.
(3) It is less sensitive to noise than more
techniques.
GFCM ALGORITHM
Step 1: Get the image and enhance the image using
contrast stretching.
Step 2: Get the number of clusters to be formed.
Step 3: Re size the image
Step 4: Resize the two dimensional image into
one dimensional array of lengthr x c
Step 5: Find the intensity range of the image
Range = [(Maximum intensity value)-

www.ijaert.org

International Journal of Advanced Engineering Research and Technology (IJAERT) 50

Volume 4 Issue 3, March 2016, ISSN No.: 2348 8190

(Minimum intensity value)]

Step 6: Find the centroid value
Centroid 1 = Range/Number of clusters
Centroid 2 = (2 Centroid 1)
Centroid 3 = (3 Centroid 1)
Centroid n = (n Centroid 1)
Step 7: Find the difference between the first
intensity value and the various centroid values.
Step 8: Based on the minimum difference obtained,
group the intensity values into the corresponding
clusters.
Step 9: Repeat step 7& 8 for all the other intensity
values of the image.
Step 10: Count the number of non zero values
present in each clusters.
Step 11: Find the mean value of each clusters and
compare with its corresponding centroid. If it is
equal then terminate else continue.
Step 12: Reshape the one dimensional array into two
dimensional images.

Aggregate all output

Defuzzification
Step 7: Finally display white matter , gray matter
and lesion part of input image in separate cluster.
Step 8: Then calculate the accuracy for lesion
detection.

5. NEURO-FUZZY INFERENCE MODELS

Neuro-Fuzzy
following types,

Inference

NFI ALGORITHM
Step 1: Get an input image and enhance it by using
Contrast Stretching.
Step 2: Resize the input image.
Step 3: Get size of image for cluster formation.
Step 4: Assign no. of cluster with their initial centroid.
Step 5: Find max & min value for centroid
assignment.
Step 6: Apply the Neuro Fuzzy Inference system.
Fuzzification
Apply Fuzzy operator
Apply Implication method

includes

5.1 MAMDANI MODEL

It was proposed in 1975. Fuzzy algorithms for
complex systems and decision processes are.
Layer-1 (Input layer)
No computation is done in this layer. Each
node in this layer, which corresponds to one input
variable, only transmits input value to the next layer
directly. The link weight in layer 1 is unity.

LAYER 5 Rule
Inference &
Defuzzification

4. PROPOSED METHOD
A fuzzy inference system can utilise human
experts by saving its essential components in rule base
and database, and perform fuzzy reasoning to derive
the output value. The derivation of if-then rules and
corresponding membership functions depends heavily
on the prior knowledge about the system under
consideration.
However there is no systematic way to
transform experiences of knowledge of human experts
to the knowledge base of fuzzy inference system. On
the other hand, neural network learning mechanism
does not rely on human expertise.
Due to homogeneous structure of neural
network, it is hard to extract structured knowledge
from either weights or configuration of the network.
The weights of the neural network represent the
coefficients of the hyper-plane that partition the input
space into two regions with different output values.

Model

LAYER 4- Rule
Consequent
layer

X1

X2

LAYER 3- Rule
antecedent layer

LAYER2Fuzzification
layer

LAYER 1- Input
layer
Figure: Architecture of Mamdani neuro-fuzzy system

www.ijaert.org

International Journal of Advanced Engineering Research and Technology (IJAERT) 51

Volume 4 Issue 3, March 2016, ISSN No.: 2348 8190

Layer-2 (Fuzzification Layer)

Each node in this layer corresponds to one
linguistic label (excellent, good, etc) to one of the
input variables in layer-1.
In other words, the output link represents the
membership value, which specifies the degree to
which an input value belongs to fuzzy set, is
calculated in layer-2.
A clustering algorithm will decide the initial
number and type of membership functions to be
allocated to each of the input variable. The final
shapes of the MFs will be fine-tuned during network
learning.
Layer-3 (Rule Antecedent Layer)
A node in this layer represents the antecedent
part of a rule. Usually a T-norm operator is used in this
node. The output of a layer-3 node represents the
Ring strength of the corresponding fuzzy rule.
Layer-4 (Rule Consequent Layer)
This node basically has two tasks. To combine
the incoming rule antecedents and determine the
degree to which they belong to the output linguistic
label (high, medium, low, etc). The number of
nodes in this layer will be equal to the number of rules.

taken as the weighted average of each rules output

thus avoiding time consuming process of
defuzzification. However this model is not used
often.

6. EXPERIMENTAL RESULTS

MRI Image ( Input Image)

Output of GFCM

Layer-5 (Combination and Defuzzification layer)

This node does the combination of all the
rules consequents using T-conorm operator and finally
computes the crisp output after defuzzification.
5.2 TAKAGI- SUGENO METHOD:
Introduced in 1985.Similar to mamdani model.
Takagi-Sugeno neuro-fuzzy systems make use of a
mixture of back- propagation to study the
membership functions and least mean square
estimation to determine the co-efficients of the linear
combinations in the rules conclusions. A step in the
studying procedure got two parts: In the first part the
input patterns are propagated, and the optimal
conclusion parameters are estimated by an iterative
least mean square procedure, while the antecedent
parameters to be fixed for the current cycle through
the training set. In the second part the patterns are
propagated repeatedly, and in this epoch, backpropagation is used to modify the antecedent
parameters are remain fixed.
5.3 TSUKAMOTO METHOD:
Avoiding time consuming process of
defuzzification. However this model is not used
often. In the Tsukamoto model, the consequent of
each fuzzy if-then rule is expressed by a fuzzy set
with a monotonical MF. As a result, the inferred
output of each rule is defined as a crisp value induced
by the rules firing strength. The overall output is

Output Image of ANFI

7. TABULATION
WML Detection and accuracy (%) of GFCM and
NFIS
Model
Accuracy
GFCM
NFIS

90.17
98.60

8. CONCLUSION
The Existing Geostatistical Fuzzy C-means clustering
(GFCM) method is used for automatic detection of
WMLs in brains of aged people. The incorporation of
the geostatistical estimate using objective functions of
fuzzy clustering and possibility, clustering algorithms
is relatively a simple and effective procedure for
implementations. Experimental results display the
advantages that Geostatistical Fuzzy C-means
clustering is a very effective approach for extracting
White Matter Lesions of about 90.17% accurate. The
demerit is that it will detect some false lesions also and
slight amount of noise will be appear. So Neuro-Fuzzy
Inference System (NFIS) algorithm is used to detect
White Matter Lesions of about 98.60% accurate which

www.ijaert.org

International Journal of Advanced Engineering Research and Technology (IJAERT) 52

Volume 4 Issue 3, March 2016, ISSN No.: 2348 8190

is more than GFCM and noise is eliminated using

trimmed filters.

9. FUTURE WORK
Future work includes the implementation of Particle
Swarm Optimization. Particle Swarm Optimization
(PSO) is a biologically inspired computational search
and optimization method developed related on the
social behaviors of birds flocking or fish schooling.

REFERENCES
[1] Scheltens P, Erkinjuntti T, Leys D, Wahlund, LO, Inzitari D, Del Ser T, Pasquier F, Barkhof F,
Mantyla R, Bowler J, Wallin A, Ghika J, Fazekas F,
Pantoni L. White matter changes on CT and MRI: an
overview of visual rating scales. Eur Neurol. 1998.
[2] L. O. Wahlund, F. Barkhof, F. Fazekas, L.Bronge,
M. Augustin,M. Sj ogren, A. Wallin,H. Ader, D.
Leys, L. Pantoni, F. Pasquier, T.Erkinjuntti, and P.
Scheltens, A new rating scale for age-related white
matter changes applicable to MRI and CT, Stroke,
vol. 32, pp.13181322, 2001
[3] Y. Wu, S. K. Warfield, I. L. Tan, W. M. Wells III,
D. S. Meier, R. A. van Schijndel, F.Barkhof, and C.
R. G. Guttmann, Automated segmentation of
multiple sclerosis lesion subtypes with multichannel
MRI, NeuroImage, vol. 32, pp. 1205 1215, 2006
[4] Khayati R., Vafadust M., Towhidkhah F., Nabavi
S. M. A novel method for automatic determination
of different stages of multiple sclerosis lesions in
brain MR FLAIR images, Computerized Medical
Imaging and Graphics. 2008;32:124133.
[5] M. delC.V.Hernendez, K. J. Ferguson, F. M.
Chappell, and J.M.Wardlaw, Newmultispectral MRI
data fusion technique for white matter lesion
segmentation: method and comparison with
thresholding in FLAIR images, Eur. Radiol., Feb. 16,
2010.
[6]
M.Anitha,
Prof.P.Tamije
Selvy
and
Dr.V.Palanisamy Automated Detection Of White
Matter Lesions In MRI Brain Images Using
Spatiofuzzy And Spatio-Possibilistic Clustering
Models Computer Science & Engineering: An
International Journal (CSEIJ), Vol.2, No.2, April
2012.
[7] D.Sushma Deevi, G.S.Ajay K Reddy, Narendra
Babu "Geostatistical And Fuzzy C-Mean Clustering
For
Extraction Of White Matter International
Journal Of Innovative Technology And Exploring
Engineering (IJITEE) ISSN: 2278-3075, Volume-2,
Issue-4, March 2013.

www.ijaert.org