Download FULL TEXT - RS Publication

International Journal of Computer Application (2250-1797)
Volume 5– No. 3, April 2015
Neuro brain MRI anatomical labeling structures segmentation using
adaptive interactive algorithm
B.Balakumar1, P.Raviraj2
1
Assistant Professor, Centre for Information Technology and Engineering, M.S University,
Tirunelveli, India, [email protected]
2
Professor, Department of CSE, Kalaignar Karunanidhi Institute of Technology, Coimbatore, Tamilnadu,
India, [email protected]
Abstract:This paper presents the primary objective of the segmentation of magnetic resonance images
(MRI) of the brain is to correctly label certain areas of the image to highlight the brain tissues, both healthy
and pathological. In practice, however, you come across often in images suffer from various kinds of
artifacts that do fail the classification algorithms. Also the effect of noise, often present in the signal
characterizing the MR images, makes the complex segmentation methods. The proposed work takes its
value in two areas of the magnetic resonance imaging (MRI) segmentation. Proper segmentation can be
performed on images without noise, therefore examines some of the algorithms that operate in the spatial
domain and in the wavelet domain. There is also an advanced algorithm segmentation that is able to well
classify the pixels of the images.
Keywords: MRI,NMR, Fuzzy K-Means clustering, anisotropic diffusion, hypothalamus.
I. INTRODUCTION
The segmentation process [1] [2] of the MR
is obtained by the matrix U that takes the name of
images of the brain has the purpose of associating
membership of each object X in each of where you
the pixels of an image of a section of the brain in
want to partition X. The present work deals with
classes that are representative of healthy tissues
segmentation resonance imaging T1-weighted
and pathological. The problem thus posed falls
magnetic and aims to achieve a segmented
within the broader cluster analysis. Given a set X
accurate structure of DGM, in order to provide a
of vectors representing the characteristics of
useful tool by which to extract quantitative
individual objects, the clustering problem in X
measures useful to the physician in diagnosis. It
consists X in the partition into several classes
describes the principles of operation of magnetic
(clusters) in such a way that the elements of each
resonance, in order to introduce the type of
class are as more like they can between them and
images. The deals with the description of the
dissimilar from the elements of other classes. The
anatomical structures of deep gray matter. In
set X is generally represented by a matrix in which
formally introduces the problem of segmentation
the rows represent the objects (vectors of
of the image. In the model Region-Scalable Fitting
characteristics) the set that you want to partition
(RSF), which is the basis of the model with the
the columns and the individual characteristics of
entropy local focus of this paper.
the returns information about the degree of
each object. Applying a clustering algorithm set X
141
International Journal of Computer Application (2250-1797)
Volume 5– No. 3, April 2015
Magnetic Resonance Imaging (MRI) is a
Spin is a fundamental property of nature, together
technique used mainly in the medical field to
with the mass and charge electricity, and is the
generate high-resolution images of the interior of
angular momentum (or magnetic) intrinsic to a
the body human, without the use of ionizing
particle. NMR is observable only the nuclei that
radiation. MRI is based on the principles of
have a nuclear magnetic moment spin, and
Nuclear Magnetic Resonance (NMR), producing
therefore behave like the needle of a compass that
images on the basis of variations, in phase and in
is oriented in an applied magnetic field. The
frequency, energy in the range of radio frequencies
nuclear magnetic moment of spin is given by the
(RF) absorbed and emitted by the object examined.
report. In which it is observed that, with
increasing intensity of the magnetic field,
increases the frequency of the precession motion,
and then the energy difference between the two
levels. The energy of a photon which needed to
cause a transition in the particle state energy
(resonance condition) is given by the relation The
objective of this work is classified, without
professional intervention, the image in white
matter (WM), gray matter (GM), cerebro-spinal
fluid (CSF), blood vessels (BV), tumor and edema
Fig 1. Brain surface segmentation in MRI
images
(Figure 1). The white matter is distributed within
Initially this technique has been designed for
located gray matter; cerebro-spinal fluid, by its
the tomographic imaging, namely for reproduction
part, flows into the cerebral ventricles, the
of a thin slice of the human body from the NMR
subarachnoid space and the ependymal canal [13].
signal, and only in Following the technique has
Image clustering of various malignant brain tumor
been extended to the imaging volume. The useful
tissues of microscopic range in H&E Stain are
signal
nuclei
analyzed through Euclidean distance metric and
characterized by non-zero spin, subject to intense
Fuzzy K-Means clustering and this method can be
magnetic fields placed in a resonant condition. To
used for pattern recognition because it provides a
understand the applications of MRI in this chapter
good object separation. Developing such tools for
will illustrate the physical base of nuclear
characterizing and effectively classifying texture
magnetic resonance, and as the MR signal can be
patterns in medical images would greatly assist in
experimentally manipulated.
the interpretation of clinical images and in
comes
from
the
hydrogen
the cortex, while that on the outside thereof is
investigating the relationship between descriptors
142
International Journal of Computer Application (2250-1797)
Volume 5– No. 3, April 2015
of texture and branching patterns and the
classification tumor region is extracted from those
associations between morphology and function or
images which are classified as malignant using
pathology[15].Manual
brain
two stage segmentation process. Experiments
tumors by medical practitioners is a time
have revealed that the technique was more robust
consuming task and has inability to assist in
to
accurate diagnosis. Several automatic methods
proposed method consists of four stages namely
have been developed to overcome these issues.
Preprocessing,
But
reduction and classification. In the first stage
Automatic
Imaging)
brain
segmentation
MRI
(Magnetic
resonance
extraction,
feature
complicated task due to the variance and intricacy
to make the image suitable for extracting the
of tumors to over by this problem we have
features. In the second stage, Region growing base
developed
automatic
segmentation is used for partitioning the image
classification of brain tumor. In the proposed
into meaningful regions. In the third stage,
method the MRI Brain image classification of
combined edge and Texture based features are
tumors is done based on Fluid vector flow and
extracted using Histogram and Gray Level Co-
support vector machine classifier. In this method
occurrence Matrix (GLCM) from the segmented
Fluid Vector Flow is utilized for segmentation of
image. In the next stage PCA is used to reduce the
two dimensional brain tumor MR images to
dimensionality of the Feature space which results
extract the tumor and that tumor can be projected
in a more efficient and accurate classification.
into the three dimensional plane to analyze the
Finally, in the classification stage, a supervised
depth of the tumor. Finally, Support vector
Radial Basics Function (RBF) classifier is used to
machine classifier is utilized to perform two
classify the experimental images into normal and
functions. The first is to differentiate between
abnormal.
normal and abnormal. The second function is to
evaluated using the metrics sensitivity, specificity
classify the type of abnormality in benign or
and accuracy. For comparison, the performance of
malignant
system
the proposed technique has significantly improved
consists of multiple phases. First phase consists of
the tumor detection accuracy with other neural
Preprocessing and segmentation, the second
network based classifier SVM, FFNN and FSVM
phased consists of first order and second order
[19].
method
tumor[16].The
for
Proposed
is
feature
anisotropic filter is applied for noise reduction and
new
segmentation
initialization, faster, and precise[17].The
a
a
tumor
of
GLCM (Gray level Co-occurrence Matrix) based
features extraction from segmented brain MR
The
obtained
experimental
are
II. SEGMENTATION BASED ON THE EXTRACTION
OF EDGES
images. Third phase classify brain images into
The phase encoding gradient is a gradient of
tumor and non-tumors using Feed Forwarded
the magnetic field Bo, used for imparting to the
Artificial neural network based classifier. After
transverse magnetization vector a specific phase
143
International Journal of Computer Application (2250-1797)
Volume 5– No. 3, April 2015
angle. The phase angle depends on the location, at
characteristic of certain discriminant analysis, but
a
transverse
has expanded into this work, incorporating the
magnetization vector. While the phase encoding
scope of analysis of accounting profitability ratios
gradient is on, each transverse magnetization
with the idea of the thick border (thick), from
vector has its own (unique) frequency. If the
studies own efficiency of financial economics. So
gradient in the X direction is switched off, the field
you will get a better discrimination between the
external magnets suffered by each spin is, for all
(profitable) companies more efficient and less,
practical
the
while that isolates industry more than possible
frequency of each transverse magnetization vector
effect (sector), which is an advantage manifest of
is identical. The phase angle φ of each carrier, on
this approach (10)
given
time
instant,
purposes,
of
identical.
the
Therefore,
the other hand, is not the same. The phase angle is
the angle that the magnetization vector form with a
III. METHODOLOGY
The brain and the brain stem are very complex
reference axis, said Y-axis, the time in which the
structures, delegated to the regulation and control
phase encoding gradient is off: the evaluation of
of vital functions of the human organism. The
the various phases of the spins then allows us to
central nervous system (CNS) consists of the
distinguish in their position along the axis X.
brain, enclosed inside the skull (from the
Encephalon, "inside the head"), and spinal cord,
content in the spinal canal. The three subdivisions
primary of the brain is the brain stem, the
cerebellum and the brain. The brainstem consists
of the bulb, the bridge and the midbrain. It is the
portion brain along which pass all the nerve fibers
that carry the signals and afferent input those
outputs efferent somatic or autonomic, between
the spinal cord and the higher cerebral centers. The
brain stem contains the cell bodies of motor
Fig 2 subcortical brain tissue segmentation
cortical brain tissue segmentation
neurons that control the skeletal muscles of the
The four components explain more than 85% of
parasympathetic, the vague (which innervates the
the original built-in variables (ratios) variance. It is
heart, smooth muscle and glands of most of the
with these new variables with we will investigate
thoracic and abdominal organs) as well as gives
the characteristics that differentiate accounting
rise to the parasympathetic fibers that innervate
firms, most and least profitable of Andalusia. It
structures of the head.
head. It also gives rise to one of the main nerves
will be, therefore, a type polar extreme approach,
144
International Journal of Computer Application (2250-1797)
Volume 5– No. 3, April 2015
(2)
The differential is an operator of the second order,
defined as the sum of the partial derivatives with
respect to the second non-mixed coordinates:
(3)
Fig 3. Anatomical segmentation in neuro MRI
Calculating the first derivative of the function f (x)
The brain stem also receives many afferent
and to zero are the coordinates of the local
fibers from the head and the visceral cavity. Along
maximum points. Effecting again the same
the entire brainstem is a structural component
operation on these coordinates, we identify the
located more inwardly, said reticular formation,
points in which the gradient turns out to be
which is constituted by a set of spread small multi-
maximized. In the field of discrete, since it
branched neurons. The neurons of the reticular
operates
formation receive and integrate information from
approximations and becomes:
on
digital
images,
you
make
several streets afferent, as well as several other
(4)
regions of the brain. Some of these neurons are
grouped together to form some of the nuclei and
To segment the DIR and FLAIR images using
"centers". The anisotropic diffusion is a technique
the model parametric active contours. First the
that eliminates the smooth out the noise of an
snake was made to evolve in the image. To further
image without its significant parts, as for example
highlight the contours of the image are applied to
the edges. The anisotropic diffusion is the result of
filter using anisotropic diffusion adapted to the
the convolution between the image and a Gaussian
gradient of the image.
filter.
IV. RESULT
The output of the reticular formation can be
(1)
Where is the, ∇ is the gradient, div (∙) is the
divergence operator and c (x, y, t) is the diffusion
coefficient that controls the speed of diffusion.
The gradient is the vector whose components
are the partial derivatives in different.
divided, from the point of functionally, in a system
descending and ascending one. The descendant’s
components affect the function of neurons both
somatic and autonomous, and the components
affecting the ascending phenomena such as the
vigil and the direction of attention on specific
events. The cerebellum has the main role of the
unconscious coordination, muscle movements and
145
International Journal of Computer Application (2250-1797)
Volume 5– No. 3, April 2015
plays a minor part in the interaction between
cord,
that
are
intended
to
terminate
in
autonomous functions and somatic functions. The
correspondence of the cerebral cortex; beams
remaining large portion of the brain is called a
descendants, or the origin of a part of the beams
brain. Its outer portion, the cortex, is a mantle cell
descendants of the spinal cord intended to go
thickness of about 3 mm, which contains about 14
directly to the suburbs or to connect with the gray
billion neurons and that covers the entire surface
matter the spinal cord. Are to be considered, in
of the brain. The bark from the outer edge of cross-
addition, the numerous Association bundles that
section of the brain. The cortex is divided into
connect the various parts of the brain.
various regions or lobes: frontal, parietal, occipital
and temporal. The cortex is an area of gray matter,
as predominantly in it neuronal cell bodies from
which originate the nerve fibers. These determine
the white matter fibers, which owes its name to the
color whitish myelin sheath of nerve fibers of the
cords.
Fig 5. Results of the segmentation
The subcortical nuclei form a zone of gray
substance which is located below the surface
formed by the bark. Their function is to contribute
to the coordination of muscle movements. The
hypothalamus is the most important area of single
control
for the
regulation
of the
internal
environment. The neurons of the hypothalamus are
also influenced by a whole set of hormones and
Fig 4. Extraction and Analysis of Segmentation
other factors chemicals circulating.
In cerebellum and in the brain white matter is
located under the bark, composed of gray matter,
more externally with respect to the nuclei (groups
V. DISCUSSION AND CONCLUSION
Magnetic resonance imaging has shown its
of neurons which constitute a part of the gray
potential for research multiple sclerosis since the
matter of the brain). The various portions of the
beginning of the 80s. The MR images allow you
brain that is distinct make detectable various types
to monitor the effects of care treatments;
of beams that constitute the white substance.
comparing two resonance images at a distance of
Indeed, we have bundles ascending, or the
time is can tell if a drug is able to reduce the
continuation of the beams ascending the spinal
number of injuries active or if it has stopped the
146
International Journal of Computer Application (2250-1797)
Volume 5– No. 3, April 2015
progression
of
the
damage
done
by MS
Encephalon. A parametric Fluid Vector Flow
comparative exploration, our proposed approach
is surpassed with existing research[18].
(FVF) active contour model is utilized for
automatic segmentation of tumor in brain MR
The principle of MRI is to adopt a coding space
images and the segmented tumor is visualized in
that allows to assign a location to each signal on a
three dimensions for depth analysis. Since a tumor
plane or a volume and therefore to reconstruct the
doesn’t exhibit any prior shape, delineating the
images. A scanner, commercial is mainly formed
tumor accurately is a difficult task. FVF is utilized
by elements that create static magnetic fields or
for segmentation because it can deform in all
variables in time and space, coordinated by a
directions for capturing the tumor. It also
complex control electronics. These elements. Also,
addresses the issues of limited capture range and
you can measure the atrophy, i.e. the degree of
the inability to extract complex contours with
reduction in the volume of brain structures and
acute
spinal
concavities.Segmentation
aids
in
cord,
reflected
associated
with
the
visualization of area of tumor[14]. In this
accumulation of disability. Are common in
methodology, we have fostered a striking tumor
patients with Multiple Sclerosis damages relating
revealing technique by maneuvering kernel
to the cortex and the matter lesions Gray plays a
ascertained SVM . The propositioned outlook
key role in the development of disability. Through
encompasses of preprocessing, segmentation,
conventional MR imaging can identify only a
feature
a
small part of such lesions, because the contrast is
preprocessing step, the noise is jettisoned and to
low between white matter, gray matter and
instigate the image appropriate for the ensuing
cerebro-spinal fluid. The use of DIR images
stages. In segmentation stage , the neoplasm
allowed the identification of intra-cortical lesisons
regions are dissected over region growing method.
in the liver. We define those curves deformable
In feature extraction, certain explicit feature will
models, or those surfaces, which move under the
be extorted by manipulating texture as well from
action of internal forces and external forces. These
intensity. On the classification stage, the kernel
forces are defined in such a way that the
based SVM is fabricated and smeared to training
deformable model is supported to the contours of
of support vector machine (SVM) to maneuver
the object under examination.
extraction
and
classification.
In
automatic detection of tumor in MRI images. For
REFERENCES
2. Juang, L. H., & Wu, M. N. (2010). MRI brain
lesion image detection based on color-converted K-
1. Brouwer, R. M., Hulshoff Pol, H. E., & Schnack,
means
H. G. (2010). Segmentation of MRI brain scans
43(7), 941-949.
clustering
segmentation.
Measurement,
using non-uniform partial volume densities. Neuro
image, 49(1), 467-477.
147
International Journal of Computer Application (2250-1797)
Volume 5– No. 3, April 2015
3. Caldairou, B., Passat, N., Habas, P. A.,
anatomical brain MRI segmentation combining
Studholme, C., & Rousseau, F. (2011). A non-local
label propagation and decision fusion. Neuro Image,
fuzzy segmentation method: application to brain
33(1), 115-126.
MRI. Pattern Recognition, 44(9), 1916-1927.
10. Sayah, Badredine, and Bornia Tighiouart.
4. Catana, C., van der Kouwe, A., Benner, T.,
"Brain tumour segmentation in MRI: knowledge-
Michel, C. J., Hamm, M., Fenchel, M., & Sorensen,
based
A. G. (2010). Toward implementing an MRI-based
approach." International Journal of Biomedical
PET attenuation-correction method for neurologic
Engineering and Technology 14.1 (2014): 71-89.
studies on the MR-PET brain prototype. Journal of
11. Jakubovic, R., et al. "Magnetic Resonance
Nuclear Medicine, 51(9), 1431-1438.
Imaging-based Tumour Perfusion Parameters are
5. Shasidhar, M., Raja, V. S., & Kumar, B. V.
Biomarkers Predicting Response after Radiation to
(2011, June). MRI brain image segmentation using
Brain Metastases." Clinical Oncology 26.11 (2014):
modified fuzzy c-means clustering algorithm. In
704-712.
Communication
Network
12. Wu, Wei, et al. "Brain tumor detection and
International
segmentation in a CRF (conditional random fields)
Technologies
Systems
(CSNT),
and
2011
system
region
framework
6. Stokking, R., Vincken, K. L., & Viergever, M. A.
superpixel-level features." International journal of
(2000).
computer
morphology-based
brain
assisted
pixel-pairwise
growing
Conference on (pp. 473-478). IEEE.
Automatic
with
and
radiology
affinity
and
and
surgery 9.2
segmentation (MBRASE) from MRI-T1 data. Neuro
(2014): 241-253.
Image, 12(6), 726-738.
13.Vijayakumar, B., and Ashish Chaturvedi.
7. Zhang, Y., Brady, M., & Smith, S. (2001).
Cut-Segmentation and Classification of MR Images
Segmentation of brain MR images through a hidden
using Texture Features and Feed Forward Neural
Markov random field model and the expectation-
Networks." European
maximization algorithm. Medical Imaging, IEEE
Transactions on, 20(1), 45-57.
A. (2004, September). Comparison study of clinical
MRI
brain
segmentation
of
Scientific
Research 85.3 (2012): 363-372.
14.Vijayakumar, B., and Ashish Chaturvedi. "Automatic
8. Song, T., Angelini, E. D., Mensh, B. D., & Laine,
3D
Journal
"Tumor
evaluation.
In
Brain Tumors Segmentation of MR Images using Fluid
Vector Flow and Support Vector Machine." Research
Journal of Information Technology 4 (2012).
Engineering in Medicine and Biology Society,
2004.
IEMBS'04.
26th
Annual
International
15.Vijayakumar, B., Ashish Chaturvedi, "Brain Tumor
Conference of the IEEE (Vol. 1, pp. 1671-1674).
in Three Dimenesional Magnetic Resonance Images
IEEE.
and Concavity Analysis."International Journal of
9. Heckemann, R. A., Hajnal, J. V., Aljabar, P.,
Computer Application 3.
Rueckert, D., & Hammers, A. (2006). Automatic
148
International Journal of Computer Application (2250-1797)
Volume 5– No. 3, April 2015
16.Vijayakumar, B., Ashish Chaturvedi,. "Effective
his M.Tech degree in the field of Computer and
Classification of Anaplastic Neoplasm in Huddling Stain
Information
Image by Fuzzy Clustering Method." International
Sundaranar
Journal of Scientific Research 3.
2006.Presently he is working as a Assistant Professor in
Centre
Technology
from
University,Tirunelveli
for
Information
Manonmaniam
in
the
year
Technology
17.Vijayakumar,B.,andAshishChaturvedi "Idiosyncrasy
Engineering,Manonmaniam
Dissection and Assorting of Brain MR Images
Tirunelveli.
Instigating
International journals and conferences. He is also
Kernel
Corroborated
Support
Vector
Machine." Archive Des Science.
B.,
has
published
many
University,
papers
in
pursuing for his Ph.D in Brain Tumour Segmentation in
the
18.Vijayakumar,
He
Sundaranar
&
and
Ashish
Chaturvedi.
area
of
image
anonmaniamSundaranar
processing
from
iersity,Tirunelveli,Tamilnadu.
"Abnormality segmentation and classification of brain
He is a member of Many life member of professional
MR images using Kernel based Support vector
bodies.
machine."Archive Des Science 66.4.
P.Raviraj
received
his
B.E
degree in Computer Science and
19.B.Balakumar
and
P.
Raviraj,
“Abnormality
Engineering
from
Maharaja
Segmentation and Classification of Brain MR Images
Engineering
using Combined Edge, Texture Region Features and
under Bharathiyar University in
Radial basics Function” Research Journal of Applied
Sciences, Engineering and Technology,Vol.6,Issue 21,pp
4040-4045,November,2013
college,coimbatore
2002.He obtained his in the
department of Computer Science
& Engineering, Kalaignar Karunanidhi Institute of
Technology, Coimbatore. He has published more than 30
papers in International journals and conferences. He is a
Regional Editor and Reviewer of five International
B.Balakumar received his B.Tech
Journals in Scialert journal publications in Newyork,
degree
and
USA. He is also a life member of professional bodies
Communication Engineering from
like ISTE, CSI etc. He has guiding the Ph.D. research
National
scholars in the areas of Image Processing, Pervasive &
in
Electronics
Institute
of
Technology(NIT),Hamirpur,H.P,I
Cloud computing, Datawarehousing and Robotics etc.
ndia in the year 2003.He obtained
149