Q04503100104

Ms. Navneet Kaur et al Int. Journal of Engineering Research and Applications www.ijera.com
ISSN : 2248-9622, Vol. 4, Issue 5( Version 3), May 2014, pp.100-104
www.ijera.com 100 | P a g e
A Novel Approach of Brain Tumor Detection Using Hybrid
Filtering
Ms. Navneet Kaur1
, Dr. Mamta Juneja2
M.E., C.S.E., U.I.E.T., P.U., CHD, India.
Assistant Professor, C.S.E., U.I.E.T., P.U., CHD, India.
ABSTRACT
The paper presented explains how the gradient differential plays an inseparable part in demarking the tumor in
brain. Areas that do not match with the benchmark set are skipped by the algorithm i.e. eminent entropy and
intensity which are considered as major feature of tumor identification. The picture is finally rebuilt by
evaluating regional maxima and extended maxima transformation which at the end gives us the most
impressionable part of tumor. To generate final output the algorithm takes just 3.98 seconds (as an average
statistics). At the end the proposed algorithm detects the tumor with high accuracy.
Keywords: Brain tumor, gabor filter, boundary, canny edge detection, extended maxima transform & image
model.etc.
I. INTRODUCTION
The main cause of brain tumor in the
humans is due to growth of abnormal cells that
originate from the tissues. Generally this cells are
died and replaced after sometime interval or when the
brain does not need them. The growth of abnormal
cells starts growing up and more cells joined which
results into cyst. The tumor is classify into two parts
first is primary tumor which is main cause of brain
tumor as cells generated from the brain parts leads to
tumor and other one is secondary tumor which cause
cancer from the tissues gathered from different parts
of the body. Tumor caused by brain cells is called
malignant which is very dangerous and effects the
parts of the brain function. The other tumor is benign
which does not effects the human body and can easily
be detected by naked eye as it shows outline around
the effected area and its rate of growth is very slow
than malignant tumor.
In our paper the brain tumor is detected from
the MRI image of brain and the area of tumor is
calculated of the effective area. Brain tumor detection
is a troublesome task as there are number of pixels in
the MRI image and detection of exact part of tumor
in it required skillful techniques. The property of the
benign tumor is having low intensity of pixels and
does not show deep marks but show slightly lining
outside the effective part. The first step in this
technique is to slice the picture into number of parts
and each part is analyses individually to detect the
tumor. If there is benign tumor in the brain than it
shows in some of the parts but if there is malign
tumor than it penetrates into deeper parts and
intensity of pixels is very high in the region. These
properties are very helpful to detect the tumor part in
the brain and we also look some of the similar
features of the malign tumor in the brain.
II. RELATED WORK
Tumor segmentation from magnetic
resonance (MR) images by hybrid approach helps to
detect tumor and tumor treatment by tracking the
tumor growth. Nobuyuki Otsua [1] has given a
technique of automatic threshold the picture
segmentation. Michael R. Kaus et.al [2] had made an
automatic brain tumor analysis method which was
gives precise results as compare to manual
segmentation with 3-D magnetic resonance images.
Lynn M. Fletcher-Heath et.al [3] in their research
stated that the automatic analysis method which
separate brain tumors which are non-enhancing from
cells in MR images which helps in calculating tumor
size over time. Alain Pitiot et.al [4] presented a fully
automated technique for medical figures. Djamal
Boukerroui et.al [5] Strong method to specially
analyse noisy images, within a Bayesian framework.
Kristin R. Swanson et.al [6] Inspects the aspects of
advancements in mathematical modeling of gliomas
in the study. Yuri Boykov et.al [7] gives low/high
flow graphs to determine energy in low level vision.
Stuart S. C. Burnett et.al [8] developed a deformable-
template algorithm for the semiautomatic delineation
of normal tissue structures on computed tomography
images. Weibei Dou et.al [9] proposed a framework
of fuzzy information fusion in this paper to
automatically segment tumor areas of human brain
from multispectral magnetic resonance imaging
(MRI) such as T1-weighted, T2-weighted and proton
density (PD) images. Kyungsuk (Peter) Pyun et.al
[10] has developed a multiclass image segmentation
RESEARCH ARTICLE OPEN ACCESS

method using hidden Markov Gauss mixture models
(HMGMMs) and provide examples of segmentation
of aerial images and textures. Hassan Khotanlou et.al
[11] has presented a new general method for
segmenting brain tumors in 3D magnetic resonance
images. Under the scope of this paper the basic
features of tumors demarcation such as texture
information have not been taken into consideration
while designing the algorithm. Jason J. Corso et.al
[12] presented a new method for automatic
segmentation of heterogeneous image data that takes
a step toward bridging the gap between bottom-up
affinity-based segmentation methods and top-down
generative model based approaches. T. Logeswari
et.al [13] in their paper had described a segmentation
method consisting of two phases. Sufyan Y. Ababneh
et.al [14] has proposed a new, fully automated,
content-based system is proposed for knee bone
segmentation from magnetic resonance images
(MRI). P. Narendran et.al [15] tried to segment brain
tumors, their components (edema and necrosis) and
internal structures of the brain in 3D MR images.
Sudipta Roy et.al [16] in their work introduced a
fully automatic algorithm to detect brain tumors by
using symmetry analysis. Mukesh Kumar et. Al [17]
used the texture analysis and seeded region growing
method which is based on texture of the MRI.
Although the author tried to minimize the total
execution time of this method but still it takes
minimum more than 8 seconds to provide its results.
III. PROPOSED WORK
The presented research paper constitutes of
an algorithm that was crafted using the hybrid
approach of combination of the technique of image
model, another is extended maxima transformation,
followed by regional maxima transformation. Later
on through canny edge detection algorithm the edges
of the tumor has been outlined for clear cut view of
the tumor potion. Before applying hybrid approach
onto the image the combination of filters i.e. gabor
filter and median filter are also being applied.

Step 1: Read and conversion of RGB to grey scale
image: first the input image is converted into the
greyscale and arranged in the form of matrix. The
data shows blue, red and green conversion at a
particular location. By using the imread command
image is analysed from the graphics file. Next step is
to convert a color image into the grayscale by using
either of two methods luminosity method or average
method. Average method gives less precise results as
it is taken by simply average of the colors.
Grayscale image = (R + G + B / 3)
The wavelengths of these color contributes much in
analyzing the image But due to different wavelengths
the output will be black in shape. To overcome this
we used the luminosity method to decrease the effect
of the red color by using new grayscale equation
New Grayscale image = ((0.3 * R) + (0.59 * G) +
(0.11 * B)).
Accordingly the Red has contributed 33%, Green is
59% (> red and blue colors) and Blue is just 11%.
Step 2: Processing : To get a clear image of MRI
which helps in detecting of tumor median filter is
applied to remove unwanted noise in the pictures.
After denoising the image the retain information is
used to arrange the intensity values. Than intensity
values are arranged with respect to their local values.
Step 3: Contrast Enhancement of Image: The most
important thing in detecting tumor is contrast
enhancement in grayscale image. The clarity of
image is improved by brightening up the image
which leads to the increase in intensity of pixels of
the effected area. The contrast of the picture is
increased by using imadjust function in matlab by
using (γ) gamma transformation. (γ) gamma
transformation function plays a vital role in
brightening up and darkening up the values of
gamma. These values are adjusted according to the
image quality and size of tumor in the brain.
Gamma compression is defined by following terms:
Vout = AVinγ
(1)
where A is a constant.
Step 4: Dividing into quadrants: The MRI image is
divided into four equal quadrants and each part is
analysed to calculate the minimum and maximum
value of pixels and compared to crop out the highest
intensity portion in the brain. Entropy is applied over
it by the following analysis
etp = - sum(p.*log2(p)) (2)
where p contains the histogram counts returned
from imhist.
Step 5: Formation of Matrix: the intensity of each
part is adjusted and compared after applying entropy
and highest
values giving quadrant is separated out after building
up a matrix. The most susceptible quadrant having
tumor is located out and next task is to locate the
exact tumor in the brain
Step 6: Thresholding: The temporal values is
calculated of the selected quadrant and thresholding
by upper and lower bound theorems. Average mean
square error of lower bound theorem is used to
analyse the quadrant and than upper bound is applied
over it.
Step 7: Maxima Transformation: To detect the tumor
portion with highest intensity extended and regional
maxima transformation is applied. The portions with
low intensity of pixels is neglected as there is very
rare chance of occurrence of tumor in that portion
and transformation is applied on the maximum
intensity part.
Step 8: Regional Properties: The characteristic of the
connected tumor portion is found out by using
regionprops command of matlab to check that
whether it is a tumor portion or not. The properties
like solidity is also checked to get the actual shape of
tumor generally uneven in nature to get sure of the
tumor part which is convex in nature.
Step9: Used Canny Edge Detection algorithm: In
digital image processing the canny edge detection
mechanism has been proved very useful in
identifying the object’s edges. So in this work the
advantage of canny edge detection algorithm has
taken for demarking the brain tumor.
Step 10: The outer edges of a suspicious region has
been marked so as to detect the area that is having
intensity maximum of all. Now the perceived tumor
portion is partitioned out and finally the region that
got impacted by the tumor is easily recognizable and
analyzed as well.
Step 11: Now the size of the tumor is quantified by
calculating the pixels that fall inside the edges and
are highlighted.
IV. RESULTS AND ANALYSIS
The following figure i.e. fig. 1 firstly shows
the original input image, secondly the resultant image
after applying median filter and gabor filter over the
input image. Third snapshot shows the quadrant
having tumor and after enhancing the contrast of that
quadrant so that the objects could be identified more
clearly followed by the output of canny edge detector
when applied over the quadrant. Finally at the end the
tumor present is being highlighted.

The table 1 followed by the snapshots gives
statistical description of each image. It shows the
results of 5 images of 5 different patients, in terms of
tumor size (in pixels) and execution time.
Fig 1. Snapshots of results
Table 1: RESULTS OF 10 DIFFERENT CASES
Case Name Type of Tumor
Time
Taken
(in sec.)
Size
(in
pixels)
A B C
Cumulative
Precision
(A/D)
Cumulative
Recall
(A/E)
1 Dorcas Malignant 2.54 62002 1 0 0 0.022 0.02
2 Jonquil Malignant 2.34 50626 2 0 0 0.044 0.04
3 Helli Malignant 2.59 66754 3 0 0 0.066 0.06
4 Daniel Benign 1.05 17685 4 0 0 0.088 0.08
5 Rhoda Malignant 2.17 49873 5 0 0 0.111 0.10
6 Amos Malignant 3.12 122145 6 0 0 0.133 0.12
7 Laura Wrong output 2.47 50626 0 1 0 0 0
8 Asaph Benign 1.59 28983 7 0 0 0.155 0.14
9 Wasila Benign 1.48 26588 8 0 0 0.177 0.16
10 Diana No output 0 0 0 0 1 0 0
V. CONCLUSION AND FUTURE
SCOPE
The table 1 shows that the crafted algorithm
has shown improved results than the earlier one in
terms of execution time taken and clarity in
understanding the image even for a non-technical
person. The constructed algorithm has very less
complexity as it chooses 1 quadrant out of 4, which
possesses the tumor and for that matter the quadrant
having quadrant having highest entropy, intensity and
solidity matrix is found. Moreover the data
processing will also be reduced as the algorithm has
to work on just one quadrant out of four. Now in
future scope the scholar could prepare machine
learning algorithms of non-parameterized features
and try to classify the horizons to the areas that
possess tumor and even to those areas that do not
possess tumor via using regression decision tree.
REFERENCES
[1] Nobuyuki Otsu, “A Threshold Selection
Method from Gray-Level Histograms”,
IEEE Transactions on systems, Man, and
Cybernetics, Vol. SMC-9, No. 1, January
1979.
[2] Michael R. Kaus, Simon K. Warfield, Arya
Nabavi, Peter M. Black, Ferenc A. Jolesz,
Ron Kikinis, “Automated Segmentation of
MR Images of Brain Tumors” Radiology;
218:586–591, Magnetic resonance (MR),
Volume measurement, 10.121412,
10.12143, 2001.
[3] Lynn M. Fletcher-Heath, Lawrence O.
Halla, Dmitry B. Goldgofa, F. Reed
Murtagh, “Automatic segmentation of non-
enhancing brain tumors in magnetic
resonance images” Artificial Intelligence in
Medicine 21: 43-63, Elsevier Science B.V.,
2001.
[4] Alain Pitiot, A.W. Toga, P.M.
Thompson, “Adaptive elastic segmentation
of brain MRI via shape-model-guided
evolutionary programming” IEEE

Transactions on Medical Imaging, Vol.: 21,
Issue: 8, Aug. 2002.
[5] Djamal Boukerroui, Atilla Baskurt, J. Alison
Noble, Olivier Basset, “Segmentation of
ultrasound images––multiresolution 2D and
3D algorithm based on global and local
statistics” Elsevier Science B.V,Vol. 24,
Issues 4–5, February 2003.
[6] Kristin R. Swanson, Carly Bridge, J.D.
Murray, Ellsworth C. Alvord Jr, “Virtual
and real brain tumors: using mathematical
modeling to quantify glioma growth and
invasion” Elsevier B.V,Vol. 216, Issue 1, 15
December 2003.
[7] Yuri Boykov, Vladimir Kolmogorov, “An
Experimental Comparison of Min-Cut/Max-
Flow Algorithms for Energy Minimization
in Vision” IEEE Transactions on Pattern
Analysis and Machine Intelligence, Vol. 26,
No. 9, September 2004.
[8] Stuart S. C. Burnett, George
Starkschall, Craig W. Stevens, Zhongxing
Liao, “A deformable-model approach to
semi-automatic segmentation of CT images
demonstrated by application to the spinal
canal” The International Journal of Medical
Physics Research and Practice, Medical
Physics 31, 251 (2004), 22 January 2004.
[9] Weibei Dou, Su Ruan, Yanping Chen,
Daniel Bloyet, Jean-Marc Constans, “A
framework of fuzzy information fusion for
the segmentation of brain tumor tissues on
MR images” Elsevier B.V,Vol. 25, Issue 2,
2006.
[10] Kyungsuk (Peter) Pyun, Johan Lim, Chee
Sun Won, Robert M. Gray, “Image
Segmentation Using Hidden Markov Gauss
Mixture Models” IEEE Transactions on
Image Processing, Vol. 16, No. 7, July 2007.
[11] Hassan Khotanlou, Olivier Colliot, Isabelle
Bloch “Automatic brain tumor segmentation
using symmetry analysis and deformable
models” Bu Ali Sina University and
Paristechile de France, 2008.
[12] Jason J. Corso, Eitan Sharon, ShishirDube,
Suzie El-Saden, Usha Sinha, Alan Yuille,
“Efficient Multilevel Brain Tumor
Segmentation With Integrated Bayesian
Model Classification” IEEE Transactions on
Medical Imaging, Vol. 27, No. 5, May 2008.
[13] T. Logeswari, M. Karnan, “An improved
implementation of brain tumor detection
using segmentation based on soft
computing” Journal of Cancer Research and
Experimental Oncology Vol. 2(1) pp. 006-
014, March, 2010.
[14] Sufyan Y. Ababneh, Jeff W. Prescott, Metin
N. Gurcan, “Automatic graph-cut based
segmentation of bones from knee magnetic
resonance images for osteoarthritis
research” Elsevier B.V.,Vol. 15, Issue 4,
August 2011.
[15] P. Narendran, Mr. V.K. Narendira Kumar,
Dr. K. Somasundaram, “3D Brain Tumors
and Internal Brain Structures Segmentation
in MR Images” I.J. Image, Graphics and
Signal Processing, 1, 35-43, February 2012.
[16] Sudipta Roy, Samir K. Bandyopadhyay
“Detection and Quantification of Brain
Tumor from MRI of Brain and it’s
Symmetric Analysis” International Journal
of Information and Communication
Technology Research, Vol. 2 No. 6, June
2012.
[17] Mukesh Kumar, Kamal K.Mehta “A Texture
based Tumor detection and automatic
Segmentation using Seeded Region Growing
Method” International Journal of Computer
Technology and Applications,Vol 2 (4),
855-859, August 2011.
[18] Ngah, U. K., Ooi, T. H., Sulaiman, S. N.
&Venkatachalam, P. A. (2002). Embedded
Enhancement Image Processing Techniques
on A Demarcated Seed Based Grown
Region. Proc. of Kuala Lumpur Int. Conf.
on Biomedical Engineering. 170-172.
[19] Lim, E. E., Venkatachalam, P. A., Ngah, U.
K. & Khalid, N. E. A. (1999). “Liver
Disease Diagnosis by Region Growing”.
Proceedings of International Conference on
Robotics, Vision and Parallel Processing for
Automation. 1. 38-45.
[20] Khalid, N. E. A., Venkatachalam, P. A. &
Ngah, U. K. (1999). “Diagnosis of Bone
Lesion Based on Histogram Equalization".
Proceedings of International Conference on
Robotics, Vision and Parallel Processing for
Automation. 1. 91-96.

Q04503100104

More Related Content

What's hot

Viewers also liked

Similar to Q04503100104

Recently uploaded

Q04503100104