Research A rticle
Image Analysis for MRI Based Brain Tumor Detection and
Feature Extraction Using Biologically Inspired BWT and SVM
Nilesh Bhaskarrao Bahadure,
1
Arun Kumar Ray,
1
and Har Pal Thethi
2
1
School of Electronics Engineering, KIIT U niversity, Bhu baneswar, Odisha, I ndia
2
Department of Electronics & Telecommunication Engineering, Lovely Professional University, J alandhar, Punjab, India
Correspondence should be addressed to Nilesh Bhaskarrao Bahadure; nbahadure@gmail.com
Received 16 January 2017; Accepted 16 February 2017; Published 6 March 2017
A
cademic Editor: Guowei Wei
Copyright © Nilesh Bhaskarrao Bahadure et al. is is an open access article distributed under the Creative Commons
Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.
e segmentation, detection, and extraction of infected tumor area from magnetic resonance (MR) images are a primary concern
but a tedious and time taking task performed by radiologists or clinical experts, and their accuracy depends on their experience
only. So, the use of computer aided technology becomes very necessary to overcome these limitations. In this study, to improve the
performance and reduce the complexity involves in the medical image segmentation process, we have investigated Berkeley wavelet
transformation (BWT) based brain tumor segmentation. Furthermore, to improve the accuracy and quality rate of the support
vector machine (SVM) based classier, relevant features are extracted from each segmented tissue. e experimental results of
proposed technique have be en evaluated and validated for performance and quality analysis on magnetic resonance brain images,
based on accuracy, sensitivity, specicity, and dice similarity index coecient. e experimental results achieved .% accuracy,
.% specicity, and .% sensitivity, demonstrating the eectiveness of the proposed technique for identifying normal and
abnormal tissues from brain MR images. e experimental results also obtained an average of . dice s imilarity index coecient,
which indicates better overlap between the automated (machines) extracted tumor region with manually ext racted tumor region
by radiologists. e simulation results prove the signicance in terms of quality parameters and accuracy in comparison to state-
of-the-art techniques.
1. Introduction
In recent times, the introduction of information technology
ande-healthcaresysteminthemedicaleldhelpsclinical
experts to provide better health care to t he patient. is study
addresses the problems of segmentation of abnormal brain
tissues and normal tissues such as gray matter (GM), white
matter (WM), and cerebrospinal uid (CSF) from magnetic
resonance (MR) images using feature extraction technique
and support vector machine (SVM) classier [, ].
e tumor is basically an uncontrolled growth of can-
cerous cells in any part of the body, whereas a brain tumor
is an uncontrolled growth of cancerous cells in the brain. A
brain tumor can be benign or malignant. e benign brain
tumor has a uniformity in structure and does not contain
active (cancer) cells, whereas malignant brain tumors have
a nonuniformity (heterogeneous) in structure and contain
active cells. e gliomas and meningiomas are the examples
of low-grade tumors, classied as benign tumors and glioblas-
toma and astrocytomas are a class of high-grade tumors,
classied as malignant tumors.
According to the World Health Organization and Ameri-
can Brain Tumor Association [], the most common grading
system uses a scale from grade I to grade IV to classify benign
andmalignanttumortypes.Onthatscale,benigntumors
fall under grade I and II glioma and malignant tumors fall
undergradeIIIandIVglioma.egradeIandIIgliomaare
also called low-grade tumor type and possess a slow growth,
whereas grade III and IV are called high-grade tumor types
and possess a rapid growth of tumors. If the low-grade brain
tumor is le untreated, it is likely to develop into a high-
grade brain tumor that is a malignant brain tumor. Patients
with grade II gliomas require serial monitoring and obser-
vations by magnetic resonance imaging (MRI) or computed
Hindawi
International Journal of Biomedical Imaging
Volume 2017, Article ID 9749108, 12 pages
https://doi.org/10.1155/2017/9749108
International Journal of Biomedical Imaging
tomography (CT) scan every to months. Brain tumor
might inuence any individual at any age, and its impact on
the body may not be the same for every individual.
e benign tumors of low-grade I and II glioma are
considered to be curative under complete surgical excursion,
whereas malignant brain tumors of grade III and IV category
can be treated by radiotherapy, chemotherapy, or a combina-
tion thereof. e term malignant glioma encompasses both
grade III and IV gliomas, which is also referred to as anaplas-
tic astrocytomas. An anaplastic astrocytoma is a mid-grade
tumor that demonstrates abnormal or irregular growth and
an increased growth index compared to other low-grade
tumors. Furthermore, the most malignant form of astrocy-
toma, which is also the highest grade glioma, is the glioblas-
toma. e abnormal fast growth of blood vessels and the
presence of the necrosis (dead cells) around the tumor are
distinguished glioblastoma from all the other grades of the
tumor class. Grade IV tumor class that is glioblastoma is
always rapidly g rowing and highly malignant form of tumors
as compared to other grades of the tumors.
To detect infected tumor tissues from medical imaging
modalities, segmentation is employed. Segmentation is nec-
essary and important step in image analysis; it is a process of
separating an image into dierent regions or blocks sharing
common and identical properties, such as color, texture,
contrast, brightness, boundaries, and gray level. Brain tumor
segmentation involves the process of separating the tumor
tissues such as edema and dead cells from normal brain
tissues and solid tumors, such as WM, GM, and CSF [] with
the help of MR images or other imaging modalities [–].
In this study, dierent magnetic resonance imaging
(MRI) sequence images are employed for diagnosis, including
T-weighted MRI, T-weighted MRI, uid-attenuated inver-
sion recovery- (FLAIR) weighted MRI, and proton density-
weighted MRI. e detection of a brain tumor at an early
stage is a key issue for providing impro ved treatment. Once
a brain tumor is clinically suspected, radiological evaluation
is required to determine its location, its size, and impact on
the surrounding areas. On the basis of this information the
best therapy, surgery, radiation, or chemotherapy, is decided.
Itisevidentthatthechancesofsurvivalofatumor-infected
patient can be increased signicantly if the tumor is detected
accurately in its early stage []. As a result, the study of brain
tumors using imaging modalities has gained importance in
the radiology department.
e rest of the paper is organized as follows: Section
presents the related works, Section presents the materials
and methods with the steps used in the proposed technique,
Section presents the results and discussion, Section
presents the comparative analysis, and nally Section
contains the conclusions and future work.
2. Related Works
Medical image segmentation for detection of brain tumor
from the magnetic resonance (MR) images or from other
medical imaging modalities is a very important process for
deciding right therapy at the right time. Many techniques
have been proposed for classication of brain tumors in MR
images, most notably, fuzzy clustering means (FCM), support
vector machine (SVM), articial neural network (ANN),
knowledge-based techniques, and expect ation-maximization
(EM) algorithm technique which are some of the popular
techniques used for region based segmentation and so to
extract the i mportant information from the medical imaging
modalities. An overview and ndings of some of the recent
and prominent researches are presented here. Damodharan
andRaghavan[]havepresentedaneuralnetworkbased
technique for brain tumor detection and classication. In this
method,thequalityrateisproducedseparatelyforsegmenta-
tion of WM, GM, CSF, and tumor region and claims an accu-
racy of % using neural network based classier. Alfonse
and Salem [] have presented a technique for automatic
classication of brain tumor from MR images using an SVM-
based classier. To improve the accuracy of the classier,
features are extracted using fast Fourier transform (FFT)
and reduction of features is performed using Minimal-
Redundancy-Maximal-Relevance (MRMR) technique. is
technique has obtained an accuracy of .%.
e extraction of the brain tumor requires the separation
of the brain MR images to two regions []. One region
contains the tumor cells of the brain and the second contains
the normal brain cells []. Zanaty [] proposed a methodol-
ogy for brain tumor segmentation based on a hybrid type of
approach, combining FCM, seed region growing, and Jaccard
similarity coecient algorithm to measure segmented gray
matter and white matter tissues from MR images. is
method obtained an average segmentation score S of %
at the noise level of % and %, respectively. Ko ng et al. []
investigated automatic seg mentation of brain tissues from
MR images using discriminative clustering and future selec-
tion approach. Demirhan et al. [] presented a new tissue
segmentation algorithm using wavelets and neural networks,
which claims eective segmentation of brain MR images into
the tumor, WM, GM, edema, and CSF. Torheim et al. [],
Guo et al. [], and Yao et al. [] presented a technique which
employed texture features, wavelet transform, and SVM’s
algorithm for eective classication of dynamic contrast-
enhanced MR images, to handle the nonlinearity o f real data
and to address dierent image protocols eectively. Torheim
et al. [] also claim that their proposed technique gives better
predictions and improved clinical factors, tumor volume, and
tumor stage in comparison with rst-order statistical features.
Kumar and Vijayakumar [] introduced brain tumor
segmentation and classication based on principal compo-
nent analysis (PCA) and radial basis function (RBF) kernel
based SVM and claims similarity index of .%, overlap
fraction of %, and an extra fraction of .%. e clas-
sication accuracy to identify tumor type of this method is
% with total errors detected of .%. Sharma et al. [] have
presented a highly ecient technique which claims accuracy
of % in the classication of brain tumor from MR images.
is method is utilizing texture-primitive features with arti-
cial neural network (ANN) as segmentation and classier
tool.Cuietal.[]appliedalocalizedfuzzyclusteringwith
spatial information to form an objective of medical image
segmentation and bias eld estimation for brain MR images.
In this method, authors use Jaccard similarity index as a
International Journal of Biomedical Imaging
measurement of the segmentation accuracy and claim %
to % accuracy to segment white matter, gray matter, and
cerebrospinal uid. Wang et al. [] have presented a med-
ical image segmentation technique based on active contour
model to deal with the problem of intensity inhomogeneities
in image seg mentation. Chaddad [] has proposed a tech-
nique of automatic feature extraction for brain tumor detec-
tionbasedonGaussianmixturemodel(GMM)usingMR
images. In this method, using principal component analysis
(PCA) and wavelet based features, the performance of the
GMM feature extraction is enhanced. An accuracy of .%
for the T-weighted and T-weighted and .% for FLAIR-
weighted MR images are obtained.
Deepa and Arunadevi [] have proposed a technique of
extreme learning machine for classication of brain tumor
from D MR images. is method obtained an accuracy
of .%, the sensitivity of .%, and specicity of .%.
Sachdeva et al. [] have presented a multiclass brain
tumor classication, segmentation, and feature extraction
performed using a dataset of MR images. In this method,
authorsusedANNandthenPCA-ANNandobservedthe
increment in classication accuracy from % to %.
e above literature survey has revealed that some of the
techniques are invented to obtain segmentation only; some of
the techniques are invented to obtain feature extraction and
some of the techniques are invented to obtain classication
only. Feature extraction and reduction of feature vectors for
eective segmentation of WM, GM, CSF, and infected tumor
region and analysis on combined approach could not be
conductedinallthepublishedliterature.Moreover,onlyafew
features are extracted and therefore very low accuracy in
tumor detection has been obtained. Also, all the above liter-
atures are missing with the calculation of overlap that is dice
similarity index, which is one of the important parameters
to judge the accuracy of any brain tumor segmentation
algorithm.
In this study, we perform a combination of biologically
inspired Berkeley wavelet transformation (BWT) and SVM
as a classier tool to improve diagnostic accuracy. e cause
of this study is to extract information from the seg mented
tumor regio n and classify healthy and infected tumor tissues
foralargedatabaseofmedicalimages.Ourresultsleadto
concludethattheproposedmethodissuitabletointegrate
clinical decision support systems for primary screening and
diagnosis by the radiologists or clinical experts.
3. Materials and Methods
is section presents t he materials, the source of brain MR
image dataset, and the algorithm used to perform brain MR
tissue segmentation. Figure provides the ow diagram of the
algorithm. As test images, dierent MR images of the brain
were used, including T-weighted MR images with Repetition
Time (TR) of and Echo Time (TE) of , T-weighted
MR images with Repetition Time (TR) of and Echo
Time (TE) of , and FLAIR-weighted MR images with Rep-
etition Time (TR) of and Echo Time (TE) of . ese
test images were acquired using a Tesla Siemens Magnetom
SpectraMRmachine.etotalnumbersofslicesforallchan-
nels were , which leads to total of images at slices or
images per patient with a eld of view of mm, an interslice
gap of mm, and voxel of size . mm ×. mm ×. mm.
e proposed methodology is applied to real dataset includ-
ing brain MR images of × pixel size and was converted
into grayscale before further processing. e following sec-
tions discuss the implementation of the algorithm.
3.1. Preprocessing. e primary task of preprocessing is to
improve the quality of the MR images and make it in a form
suited for further processing by human or machine vision
system. In addition, preprocessing helps to improve certain
parameters of MR images such as improving the signal-to-
noise ratio, enhancing the visual appearance of MR image,
removing the irrelevant noise and undesired parts in the
background, smoothing the inner part of the region, and
preserving its edges []. To improve the signal-to-noise ratio,
and thus the clarity of the raw MR images, we applied adaptive
contrast enhancement based on modied sigmoid function
[].
3.2. Skull Stripping. Skull stripping is an important process in
biomedical image analysis, and it is required for the eective
examination of brain tumor from the MR images [–].
Skull stripping is the process of eliminating all nonbrain
tissues in the brain images. By skull stripping, it is possible to
remove additional cerebral tissues such as fat, skin, and skull
in the brain images. ere are several techniques available for
skullstripping;someofthepopulartechniquesareautomatic
skullstrippingusingimagecontour,skullstrippingbased
on segmentation and morphological operation, and skull
stripping based on histogram analysis or a threshold value.
Figure provides the stages of the skull stripping algorithm.
is study uses the skull stripping technique that is based on
a threshold operation to remove skull tissues.
3.3. Segmentation and Morphological Operation. e seg-
mentation of the infected brain MR regions is achieved
through the following steps: In the rst step, the preprocessed
brainMRimageisconvertedintoabinaryimagewitha
thresholdforthecut-oofbeingselected.epixelvalues
greater than the selected threshold are mapped to white,
while others are marked as black; due to this two, dierent
regions are formed around the infected tumor tissues, which
is cropped out. In the second step, in order to eliminate
white pixel, an erosion operation of morphology is employed.
Finally, the eroded region and the original image are both
divided into two equal regions and the black pixel region
extracted from t he erode operation is counted as a brain MR
image mask. In this study, Berkeley wavelet transformation is
employed for eective segmentation of brain MR image.
A wavelet is a function that is dened over a nite
intervaloftimeandhasanaveragevalueofzero.ewavelet
transformation technique is employed to develop functions,
operators, data, or information into components of dierent
frequency, which enables studying each component sepa-
rately. All wavelets are generated from a basic wavelet Ψ()
International Journal of Biomedical Imaging
Enhancement
Skull stripping
Removal of skull
Separation of GM,
WM, CSF, and tumor
Segmentation
Feature extraction
Mean, contrast,
entropy, and energy
Morphological
operation
Area extraction &
decision making
Classification using
SVM
MR image
dataset
Normal tissue
Abnormal tissue
Pre processing
F : Steps used in proposed algorithm.
Input image
Convert image to grayscale
Convert image to binary image by thresholding
Find the number of connected objects
Find mask by assigning 1 to inside and 0 to outside
of the object that show brain region
Multiply the mask with T1, T2 and FLAIR MR images
to get their skull-stripped MR image
F : Steps us ed in the skull stripping algorithm.
by using the scaling and translation process dened by (); a
basic wavelet is also referred to as a mother wavelet because
it is the point of origin for other wavelets.
Ψ
𝑠,𝜏
=
1
Ψ
−
,
()
where and are the scale and translation factors, respec-
tively.
e Berkeley wavelet transform (BWT) [, ] is
describ ed as a two-dimensional triadic wavelet transform
and can be used to process the signal or image. Just like the
mother wavelet transformation or other families of wavelet
transformation, the BWT algorithm will also perform data
conversion from a spatial form into temporal domain fre-
quency. e BWT presents an eective way of representation
of image transformation and it is a complete orthonormal
[]. e mother wavelet transformation
𝜑
𝜃
is piecewise
constant function [, ]. e substitute wavelets from the
mother wavelet
𝜑
𝜃
areproducedatvariouspixelspositionsin
the two-dimensional plane through scaling and translation of
themotherwaveletanditisshownin
𝜑
𝜃
(
,
)
=
1
2
𝜑
𝑥
3
𝑠
(
−
)
,3
𝑠
−,
()
where and are translation and scale parameter of the
wavelet transformation, respectively, and
𝜑
𝜃
is the trans-
forming function, and it is called the mother wavelet of
Berkeley wavelet transformation. e only single constant
term is sucient to represent the mean value of an image; the
coecient value of the single term is shown in
0
=
1
9
3
,
3
.
()
e morphological operation is used for the extraction
of the b oundary areas of the brain images. Conceptually,
the morphological operation is only rearranging the relative
order of pixel values, not on their mathematical values, and so
issuitabletoprocessonlybinaryimages.Dilationanderosion
are the two most basic operations of morphology. Dilation
operations are intended to add pixels to the boundary region
oftheobject,whileerosionoperationsareintendedtoremove
thepixelsfromtheboundaryregionoftheobjects.eoper-
ation of addition and removing pixels to or from boundary
region of the objects is based on the structuring element of
the selected image.
e experimented results produced by the proposed
technique depicted for the segmented outcome for the three
classes of WM, GM, and CSF and for the extracted tumor
International Journal of Biomedical Imaging
(a) (b) (c) (d) (e) (f)
(g) (h) (i) (j) (k) (l)
F : Segmented and area extracted result of brain MR image. (a) Original image. (b) Enhanced image. (c) Skull-stripped image. (d)
Wavelet transpose image. (e) Intense segmented image. (f) Inverse intense image. (g) Gray matter. (h) White matter. (i) CSF. (j) Dice overlap
image. (k) Eroded image. (l) Area extracted image.
region are given in Figure . e experimental results als o
nd dice overlap image, indicating the comparison between
the algorithm output and ground truth.
3.4. Feature Extraction. It is the process of collecting higher-
level information of an image such as shape, texture, color,
andcontrast.Infact,textureanalysisisanimportantparame-
ter of human visual perception and machine learning system.
It is used eectively to improve the accuracy of diagnosis
system by select ing prominent features. Haralick et al. []
introduced one of t he most widely used image analysis
applications of Gray Level Cooccurrence Matrix (GLCM)
andtexturefeature.istechniquefollowstwostepsfor
feature extraction from the medical images. In the rst step,
theGLCMiscomputed,andintheotherstep,thetexture
featuresbasedontheGLCMarecalculated.Duetothe
intricate struc ture of diversied tissues such as WM, GM, and
CSFinthebrainMRimages,extractionofrelevantfeatures
is an essential task. Textural ndings and analysis could
improve the diagnosis, dierent stages of the tumor (tumor
staging), and therapy response assessment. e statistics
feature formula for some of the useful features is listed below.
(1) Mean (M).emeanofanimageiscalculatedbyadding
all the pixel values of an image divided by the total number of
pixels in an image.
=
1
×
𝑚−1
𝑥=0
𝑛−1
𝑦=0
,.
()
(2) Standard Deviation (SD).estandarddeviationisthe
second central moment describing probability distribution
of an observed population and can serve as a measure of
inhomogeneity. A higher v alue indicates better intensity level
and high contrast of edges of an image.
SD
(
)
=
1
×
𝑚−1
𝑥=0
𝑛−1
𝑦=0
,−
2
.
()
(3) Entropy (E). Entropy is ca lculated to characterize t he
randomness of the textural image and is dened as
=−
𝑚−1
𝑥=0
𝑛−1
𝑦=0
,
log
2
,
.
()
(4) Skewness (
𝑘
). Skewness is a measure of symmetry or the
lack of symmetry. e skewness of a random vari able is
denoted as
𝑘
()and it is dened as
𝑘
(
)
=
1
×
∑
,−
3
SD
3
.
()
(5) Kurtosis (
𝑘
). e shape of a random variable ’s probability
distribution is described by the parameter called Kurtosis. For
the random variable , the Kurtosis is denoted as
urt
()and
it is dened as
urt
(
)
=
1
×
∑
,−
4
SD
4
.
()
(6) Energy (En). Energy can be dened as the quantiable
amount of the extent of pixel pair repetitions. Energy is a
parameter to measure the similarity of an image. If energy is