International Journal of Computer Applications (0975 – 8887)
Volume 138 – No.13, March 2016
6
Brain Tumor Segmentation
Jenish Gada
Student of BE
Department of
Computer
K. J. Somaiya
College Engineering
Mumbai, India
Akash Savla
Student of BE
Department of
Computer
K. J. Somaiya
College Engineering
Mumbai, India
Smit Chheda
Student of BE
Department of
Computer
K. J. Somaiya
College Engineering
Mumbai, India
Poonam Bhogale
Asst. Professor
Department of
Computer
K. J. Somaiya
College Engineering
Mumbai, India
ABSTRACT
Tumor is an uncontrolled growth of tissues in any part of the
body. Tumors are of different types and they have different
characteristics and different treatment. As it is known, brain
tumor is inherently serious and life threatening. Brain tumor
analysis is done by doctors but its grading gives different
conclusion which may vary from one doctor to another.
However this method of detection resists the accurate
determination of size of tumor. To avoid that, uses computer
aided method for segmentation of brain tumor based on the
combination of three algorithms. This algorithm allows the
segmentation of tumor tissue with accuracy comparable to
manual segmentation. It also reduces time analysis. At the end
of the process the tumor is extracted for MR image and its
exact position and its shape is also determined.
Keywords
Magnetic Resonance Imaging (MRI), Pre-Processing, K-
means, Fuzzy c-means, Linde- Buzo- Gray algorithm.
1. INTRODUCTION
Image Segmentation is a fundamental issue in biomedical
imaging area. The system aims at studying and comparing
segmentation algorithms, based on standard parameters like
mean human interaction, area, entropy, peak signal to noise
ratio. Our objective is to develop a system incorporating
image processing and computer vision techniques for
enhancement, segmentation of brain tumor. Our aim is
enhancing the current accuracy of diagnostic using industry
standard simulation software tool, matlab and dataset. The
system is expected to improve the efficiency of brain tumor
screening, and possibly reduce health care costs by decreasing
for follow-up procedures.
2. RELATED WORK
The existing method is based on the thresholding and region
growing. The thresholding method offered the possibility of
conducting a simple and fast segmentation but ignored the
spatial characteristics. Normally spatial characteristics are
important for the tumor detection. In the thresholding based
segmentation the image is considered as having only two
values either black or white. But in bit map image consists 0
to 255 gray scale values so sometimes it may ignore the tumor
cell also; this technique is generally used as first step in the
segmentation process. In the case of the region growing based
segmentation it needs more user interaction for selection of
the seed. Seed is nothing but the Centre of the tumor cells; it
may cause intensity in homogeneity problem. This is the main
drawbacks of existing system hence this is the proposed
method for tumor segmentation.
3. BLOCK DIAGRAM
Fig.1. Block Diagram
User need to first select the MRI image. The MRI images
which are obtained from the various sources typically have
huge size of the image due to which they are highly
susceptible to noisy and inconsistent data. Processing low
quality data will lead to low quality result. It is required to
improve image data by suppressing unwanted distortions and
enhance image for further processing. Image Pre-Processing is
nothing but normalizing intensity of the image, resizing an
image, colour to gray scale conversion and filtering the image
from noise. The next step is to apply the image for
segmentation.
4. METHODS
4.1 K-means Clustering
[5][6][7]
K-means is one of the unsupervised learning algorithms for
cluster. In the K-means algorithm initially the number of
cluster k has to be defined. Then k-cluster centre are chosen
randomly. The distance between the each pixel to each cluster
centres are calculated. Single pixel is compared to all cluster
canters using distance formula. The pixel is moved to the
particular cluster which has shortest distance among all. Then
International Journal of Computer Applications (0975 – 8887)
Volume 138 – No.13, March 2016
7
the centroid is re-estimated. The process continuous until the
centre converges.
4.1.1 K-means Algorithm
Give the number of cluster value as k.
Randomly choose the k cluster centres.
Calculate the centre of the cluster.
Calculate the distance between each pixel to each
cluster centre.
If distance is near to the centre then move to that
cluster.
Otherwise move to the next cluster.
Re-estimate the centre.
Repeat the process until the centre doesn’t move.
4.1.2 Output of K-means Clustering
Fig.2. Output of K-means Clustering
4.2 Fuzzy c-mean Clustering
[5[6][7]
In fuzzy c-mean the data has to be processed by giving the
partial membership value to each pixel in the image. The
membership value of the fuzzy set is ranges from 0 to 1. In
fuzzy clustering basically member of one fuzzy set can also be
member of other fuzzy sets in the same image. These are the
three basic features involved in characterization by member
function. The core is the fully member of the fuzzy set. The
support is non-membership value of the set and boundary is
the partial membership with the value between 0 and 1.
4.2.1 Fuzzy c-mean Algorithm
Randomly select the k cluster centre.
Calculate the fuzzy membership using distance
formula.
Calculate the fuzzy centre using formula.
Repeat steps 2 and 3 until minimum value is
achieved.
4.2.2 Output of Fuzzy c-mean
Fig.3. Output of Fuzzy c-means
4.3 Linde Buzo Gray Algorithm (Vector
Quantization)
[8]
The LBG algorithm is an iterative algorithm. Two
dimensional vector space are been created. It requires an
initial codebook to start with. Codebook is generated using a
training set of images. The training vector space is created and
then the centroid is obtained. The centroid is considered as the
first code vector. Now, constant errors are added to the code
vector and two new vectors are created. By using Euclidean
distance two clusters are obtained in first iteration. In the next
iteration same procedure is repeated.
4.3.1 LBG Algorithm
Generate the training vector space, T of the image
which contains M training vectors.
Find centroid c, this centroid is the first code vector.
Two new vectors are obtained after adding constant
error to code vector.
Find the Euclidian distance of the training vector
space with these two vectors.
Put the training vector in first cluster if the distance
between the training vector and the code vector is
less else put the training vector in the second
cluster.
Repeat the steps 2 to 5 for every cluster.
Stop when desired codebook size is obtained.
4.3.2 Output of LBG Algorithm
Fig.4. Output of LBG algorithm
5. COMPARISON
Table 1: Comparison of algorithm.
PARAMETER
KMEANS
FUZZY C
MEAN
LBG
TIME
(approx.)
8.18 sec
9.73 sec
60.22 sec
AREA
(approx.)
21.06
906.25
707
International Journal of Computer Applications (0975 – 8887)
Volume 138 – No.13, March 2016
8
Fig.5.A sample line graph of area comparison
Fig.6. A sample line graph of time comparison
The above mentioned algorithm are implemented and tested
on the basis of parameters like area and time for
approximately 20 images. As the Fuzzy c means gives the
optimum result as it gives the maximum area. The lbg
requires the more time to execute then other algorithms. The
area segmented by k means is less precise then other
algorithm.
6. CONCLUSION
The three algorithms k-means, fuzzy c-mean and lbg
algorithm were successfully implemented and also reviewed
the techniques of the MRI image enhancement in terms of
tumor pixel detection. Comparison of these algorithms are
done on the basis of time and area. This paper gives enhanced
information about brain tumor detection and segmentation.
The future scope of the project is classification of the tumor.
7. REFERENCES
[1] Brain Tumor Segmentation and Area Calculation of
Tumor by Use of Unsupervised Clustering Algorithm
Sonali Javare, Nitin Patil, Sanjay Patil ,D. N. Patel CoE,
Shahada, India, Volume 4, Issue 11, November 2014,
International Journal of Advanced Research in Computer
Science and Software Engineering..
[2] DETECTION OF BRAIN TUMOUR USING
CLASSIFICATION ALOGRITHM Amruta Chitari
and Mr V.V.BAG PG Research Scholar, N.K. Orchid
College of Engineering and Technology, Solapur,
Associate Professor, N.K. Orchid College of Engineering
and Technology, SolapurInternational Journal of
Inventions in Computer Science and Engineering ISSN,
Volume 1 Issue 6 July 2014.
[3] Segmentation of Brain Tumour and Its Area
Calculation Varsha Kshirsagar, Prof.Jagruti Panchal,
EnTC Dept & Pune University, India Volume 4, Issue 5,
May 2014 International Journal of Advanced Research
in Computer Science and Software Engineering.
[4] Comparison Study of Segmentation Techniques for
Brain Tumour Detection D. Manju1, Dr. M. Seetha2,
Dr. K. Venugopala Rao3 123Department of CSE, G
Narayanamma College for Women Shaikpet, Hyderabad,
India. IJCSMC, Vol. 2, Issue. 11, November 2013.
[5] Brain Tumour Segmentation Using K-Means And
Fuzzy C-Means Clustering Algorithm Nimeesha K M ,
Rajaram M Gowda, Asso. Professor, Department of
Information Science and Engineering, M.S.Ramaiah
Institute of Technology, International Journal of
Computer Science & Information Technology Research
Excellence Vol. 3, Issue 2, Mar.-Apr. 2013.
[6] Brain Tumor Segmentation: A Performance Analysis
using K-Means, Fuzzy C-Means and Region Growing
Algorithm Heena Hooda , Om Prakash Verma, Tripti
Singhal Department Of IT, Delhi Technological
University, INDIA 2014 IEEE International Conference
on Advanced Communication Control and Computing
Technologies.
[7] Brain Tumor Segmentation and Its Area Calculation
in Brain MR Images using K-Mean Clustering and
Fuzzy C-Mean Algorithm J.selvakumar A.Lakshmi
T.Arivoli, IEEE-International Conference On Advances
In Engineering, Science And Management (ICAESM -
2012) March 30, 31, 2012.
[8] An Improved LBG Algorithm for Image Vector
Quantization Bang Huang Institute of System
Engineering, Jiangnan University Wuxi, China 978-1-
4244-5540-9/10/$26.00 ©2010 IEEE
0
200
400
600
800
1000
K means
Fuzzy c
means
LBG
Area
Alogrithms
0
20
40
60
80
K means
Fuzzy c
means
LBG
Time (sec)
Alogrithms
IJCA
TM
: www.ijcaonline.org