Mri brain image segmentatin and classification by modified fcm &svm akorithm

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
_______________________________________________________________________________________
Volume: 02 Issue: 12 | Dec-2013, Available @ http://www.ijret.org 666
MRI BRAIN IMAGE SEGMENTATIN AND CLASSIFICATION BY
MODIFIED FCM &SVM AKORITHM
G.B.Deshmukh, P.D.Lambhate
1
M.E. Student , Computer Engineering, JSCOE, Maharashtra, India, gopaldeshmukh16@yahoo.in
2
Asso Professor, Information Technology, JSCOE, Maharashtra, India, pinu_poonam@yahoo.co.in
Abstract
Brain Tumor detection is challenging task in biomedical field. Image segmentation is a key step from the image
processing to image analysis, it occupy an important place. The manual segmentation of brain image is challenging
and time consuming task. An automated system overcomes the drawbacks as well as it segments the white matter, grey
matter, cerebrospinal fluid and edema. This clustering approach is particularly used for brain tumor detection in
abnormal MR images. In this paper the application of Modified FCM algorithm for Brain tumor detection and its
classification by SVM algorithm is focused. The Magnetic Resonance image is converted in to vector format and that
is given as input to the modified fuzzy c-means algorithm. In modified fuzzy c-means the steps are: initial fuzzy
partitioning and fuzzy membership generation Cluster updation based on objective function, Assigning labels to pixels
of each category and display segmented image that will give more meaningful regions to analyze. This clustered
images served as inputs to SVM. The basic SVM takes a set of input data and predicts, for each given input, which of
two possible classes.
Keywords: Clustering, Classification, Fuzz C-Means, Support Vector Machine, MRI, Brain Tumor.
--------------------------------------------------------------------***----------------------------------------------------------------------
1. INTRODUCTION
The medical field is very vast and challenging. Medical
image analysis has great importance especially in non-
invasive treatment and study of deceases. Analysis tool
helps both doctor and radiologist to arrive at specific
diagnosis of disorder. Medical image data is obtained form
magnetic resonance immerging (MRI), Computed
Tomography(CT) etc which indicates the presence or
absence of lesion along with patient history. The
segmentation technique mostly preferred by Radiologist to
segment input medical image intends the size of tumor can
be tracked using this technique which helps for treatment
planning. Previously manual segmentations used which is
time consuming and prone to human errors. Challenges in
Automatic Brain MR Image .
1.1 Tumor Segmentation
One of the most prominent issues in brain tumor
segmentation is that tumor pixels could have similar or
identical signal intensity in comparison to normal pixels
even within the same image. An additional challenge is that
tumor areas often have heterogeneous intensities.
Furthermore, the large variation of properties amongst
different tumor types makes the task of distinguishing tumor
from normal brain tissues more complex. Meaningful
regions. The abnormal MR image is segmented so as to
detect Tumor. The specific area Detection of brain tumor
can be done with the help of segmentation but to classify its
type is challenging process. This need motivates to combine
both clustering and classification together. Image
segmentation can be performed effectively by clustering
image pixels. Modified Fuzzy C-Means is used to clustering
and Support Vector machine algorithm is used for
classification.
2. RELATED WORK
A researcher in the field of biomedical engineering not only
needs to be familiar with the relevant applications of
engineering in medicine but also with the basic life sciences
[1].This interaction between the traditional engineering field
and modern medicine is for us the motivation to tackle this
paper. It is interesting to see how new techniques in the field
of engineering can improve medical diagnosis and health
care. Many techniques have been reported for clustering and
classification of brain tumor in MR images. Clustering is
one of the widely used segmentation technique. Fuzzy C-
Means algorithm is mostly preferred due to its flexibility
.But major operational drawback is that it is time consuming
[2]. A hierarchical FCM algorithm uses the concept of
template matching proposed by Kwon and Han[4].But has
drawback of requirement of accurate emplate.FCM is also
implemented by parallel processing [3] but hardware
implementation is not effective it will not achieve desired
efficiency. Cheng and Goldgof [4] proposed the fast
clustering algorithm based on random sampling. Fast fuzzy
clustering for web documentation which is highly robust is
proposed in [5] .Modified Fuzzy C-Means algorithm uses
the technique of data compression for clustering [6].But has
the drawback it is not able to specify the class of tumor.
Many techniques have been reported for classification of

_______________________________________________________________________________________
brain tumors in MR images, most notably, support vector
machine (SVM) [7] neural network [8], knowledge based
techniques [9].Classification rate of Neural Network is
lower than SVM Chaplot etal [10].SVM it is an very
powerful method of obtaining model for classification
[11].SVM provides a mechanism for choosing model
structure in which it gives low generalizationrisk.SVM
produces output which is combination of the training
example projected on high dimensional feature space
through the use of kernel space. been reported for clustering
and classification of brain tumor in MR images. Clustering
is one of the widely used segmentation technique. Fuzzy C-
Means algorithm is mostly preferred due to its flexibility
.But major operational drawback is that it is time consuming
[2]. A hierarchical FCM algorithm uses the concept of
template matching proposed by Kwon and Han[4].But has
drawback of requirement of accurate template.FCM is also
implemented by parallel processing [3] but hardware
implementation is not effective it will not achieve desired
efficiency. Cheng and Goldgof [4] proposed the fast
clustering algorithm based on random sampling. Fast Fuzzy
clustering for web documentation which is highly robust is
proposed in [5] .Modified Fuzzy C-Means algorithm uses
the technique of data compression for clustering [6].But has
the drawback it is not able to specify the class of tumor.
Many techniques have been reported for classification of
brain tumors in MR images, most notably, support vector
machine (SVM) [7] neural network [8], knowledge based
techniques [9].Classification rate of Neural Network is
lower than SVM Chaplot etal [10].SVM it is an very
powerful method of obtaining model for classification
[11].SVM provides a mechanism for choosing model
structure in which it gives low generalizationrisk.SVM
produces output which is combination of the training
example projected on high dimensional feature space
through the use of kernel space.
Fig -1: MRI Image1
Fig-2:MRI Image2
3. PROPOSED WORK
In our proposed method, we have to pre-process the given
test image for reducing noise and to enhance the contrast.
Then, we have to segment the brain image using modified
fuzzy c-means algorithm to visually show the normality’s
such as tumor. Afterwards, texture features (GLCM) will be
extracted from it. In feature extraction stage, statistical
measurements are calculated from the gray level co-
occurrence matrix for different directions and distances.
Among the various features extracted. We have to select the
distinct features that will be utilized for classification
purpose. For the selection of features SFS(Sequential
Forward Selection) is used. Support vector machine (SVM)
is used to classify whether the test image comes under
normal, benign and malignant.
3.1 Steps For Tumor Classification
1. Preprocessing (Median filter)
2. Segmentation (Modified FCM)
3. Feature extraction (GLCM technique)
4. Feature selection (Sequential forward selection)
5. Classification by SVM
1. Preprocessing (Median filter)
1. Median Filtering:
Median filtering operation, the pixel values in the
neighborhood window are ranked according to intensity, and
the middle value (the median) becomes the output value for
the pixel under evaluation. Median filtering does not shift
boundaries, as can happen with conventional smoothing
filters. Since the median is less sensitive than the mean to
extreme values (outliers), those extreme values are more
effectively removed. Median filtering preserves the edges.
2. Segmentation (Modified FCM)
First the conventional FCM was analyzed. The results show
that it does not produce a standard segmentation result
always due to the random nature of initial membership
values. This is the main drawback in conventional FCM. In
this phase modified fuzzy c-means segmentation is
implemented to overcome this problem by initializing
standard membership values (computed by template
averaging) so that standard segmentation result for each and
every image was obtained. The number of clusters, initial
standard fuzzy membership values and the feature vector
were given as the input. After the completion of initial
clustering process, the clusters were updated in turn
updates the cluster weights and membership degree. An
objective function will determine that up to which the
clusters should be updated. Vector membership values are
distributed identically to all members in that cluster. The
convergence rate is highly improved when compared with
the conventional FCM. Finally the intra cluster similarity
was computed to determine which cluster is having high
similarity among its members. In addition to that the
algorithm was executed for various numbers of clusters for
the computation of Davies Bouldin index. This index was
useful in the prediction of correct number of clusters to be
given in advance. Fuzzy partitioning is carried out through

_______________________________________________________________________________________
an iterative procedure that updates membership uij and the
cluster centroids cj by
where m > 1, and represents the degree of fuzziness
(typically=2)
3. Feature extraction (GLCM technique)
Extraction of Text Feature:-
Text features are extracted by using Gray Level Co-
occurrence Metrics Technique. A GLCM Pd[i,j] is defined
by first specifying a displacement vector d=(dx,dy) and
counting all pairs of pixels separated by d having gray levels
i and j.
The GLCM is defined by:– where nij is the number of
occurrences of the pixel values (i,j) lying at distance d in the
image.
The co-occurrence matrix Pd has dimension n× n, where n
is the number of gray levels in the image.
4. Feature selection (Sequential forward selection)
Automatic feature selection is an optimization technique
that, given a set of features, attempts to select a subset
of size that leads to the maximization of some criterion
function. Feature selection algorithms are important to
recognition and classification systems because, if a feature
space with a large dimension is used, the performance of the
classifier will decrease with respect to execution time and to
recognition rate. The execution time increases with the
number of features because of the measurement cost. The
recognition rate can decrease because of redundant features
and of the fact that small number of features can alleviate
the course of dimensionality when the training samples set is
limited, leading to overtraining. On the other hand, a
reduction in the number of features may lead to a loss in the
discrimination power and thereby lower the accuracy of the
recognition system. In order to determine the best feature
subset for some criterion, some automatic feature selection
algorithm can be applied to the complete feature space,
varying the number of selected features from to .
5. Classification by SVM
1) Training/Clustering Stage: the process of defining
criteria by which patterns are recognized, developing a
numerical description for each class.
2) Classification Stage: each pixel in the image data set
is categorized into the class it most closely resembles
based on a mathematical decision rule
3) Output Stage: results are presented in a variety of
forms (tables, graphics, etc.)
Fig 3: Proposed Design
CONCLUSION
The work in this paper involves Modified Fuzzy C-
Means for segmentation the input which is MR Image
to detection of abnormality from brain and SVM to
classify the input from modified FCM to detect the
type of tumor. The intention of the paper is to get
More accurate classification of tumor into four types.
REFERENCES
1
2
1 ||||
||||
1


 








mC
k ki
ji
ij
cx
cx
u




 N
i
m
ij
N
i
i
m
ij
j
u
xu
c
1
1
Feature Vector
Training Set Test Image
Preprocessing
(Median
Filter)
Preprocessing
(Median Filter)
Segmentation
(Modified
FCM)
Segmentation
(Modified FCM)
Feature
Extraction(GLCM)
Feature
Extraction(GLCM)
Feature
Selection(SFS)
Feature Subset
Feature Subset
Matrix
SVM-Training
SVM-
Structure
SVM-Testing
Classification
Result

_______________________________________________________________________________________
[1]Breitenmoser,”Evalutation and implementation of neural
brain activity detection methods for fMRI”, Linkoping
University, Sweden,21 February 2005,pp.13-18.
[2] Jie Yu, Peihuang Guo, Pinxiang Chen, Zhongshan Zhang
and Wenbin Ruan, ”Remote sensing image classification
based on improved fuzzy c-means”, Geo-Spatial
Information Science,IEEE transactionsvo1.11, no.2, pp:90-
94, 2008.
[3] S.Murugavalli and V.Rajamani, ”A high speed
parallelfuzzy Cmean algorithm for brain tumor
segmentation”, BlME journal,IEEE transactions vo1.6, no.1,
2006
[4] Tai Wai Cheng, Goldgof and Lawrence Hall, ”Fast fuzzy
clustering”, Fuzzy Sets and Systems,IEEE transactions
vo1.93, pp:49-56, 1998.
[5] R Krishnapmam and Y Li, ”A fuzzy relative of the k-
medoids algorithm with application to web document and
snippet clustering”, IF-EE nternationalFuzzy Systems
conference, vo1.3 pp:1281-1286, 1999.
[6] M. Shasidhar, V.Sudheer Raja, B. jayKkumarmri brain
image segmentation using modified fuzzy cmeans clustering
algorithm , 978-0-7695-4437-3/11 , 2011 IEEE.
[7] Mohd Fauzi Bin Othman, Noramalina, Bt Abdullah,
Nurul Fazrena, Bt Kamal MRI Brain classification using
Support vector machine 978-1-4577-0005- 7/11/,2011 IEEE.
[8] Shan Shen, William Sandham, Member, IEEE, Malcolm
Granat, and Annette Sterr mri fuzzy segmentation of brain
tissue using neighborhood attractionwith neural-network
optimization IEEE transactions on information technology
in biomedicine, vol. 9, no. 3, september 2005
[9] Clark M.C., Hall L.O., Goldgof D.B., Velthuizen R.,
Murtagh F.R. and Silbiger M.S.Automatic tumor
segmentation using knowledge based techniques, IEEE
Transactions on Medical Imaging, 1998, 17(2),p. 187-
192.[10] Chaplot S., Patnaik L.M., Jagannathan N.R.,
Classificationof magnetic resonance brain imagesusing
wavelets as input to support vector machine and neural
network, Biomedical SignalProcessing and Control, 2006,
1(1), p. 86-92.7
[11] C. J. C. Burges, “A tutorial on support vector machines
for pattern recognition”,Data Mining and Knowledge
Discovery, 2(2):121–167, 1998.
[12] Shan Shen, William Sandham, Member, IEEE,
Malcolm Granat, and Annette Sterr IEEE transactions on
information technology in biomedicine, vol. 9, no. 3,
September 2005
[13] S R Kannan, ”Segmentation of MRI using new
unsupervised fuzzy C-means algorithm”, ICGST-GVIP
journal, vo1.5, no.2, 2005.
[14] S.Murugavalli and V.Rajamani, ”A high speed parallel
fuzzy Cmean algorithm for brain tumor segmentation”,
BlME journal, vo1.6, no.1, 2006.8
BIOGRAPHIES
G.B.Deshmukh is M E Computer student and currently
doing work on data mining (classification and clustering of
MRI Brain Image for Tumor.)
P.D.Lambhate is Associate professor and Head of the
department from Information Technology.

Mri brain image segmentatin and classification by modified fcm &svm akorithm

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Mri brain image segmentatin and classification by modified fcm &svm akorithm (20)

More from eSAT Journals (20)

Recently uploaded (20)