International Journal of Engineering Trends and Technology(IJETT) - Volume 4 Issue 1-Jan 2013
Hemangioblastoma Neoplasm of FLAIR
MRI Confront in Vigorous Dissection
B.Rajesh Kumar#1, Dr. K.P. Yadav*2
#
Research Scholar, Department of Computer Science and Engineering,
SunRise University, Alwar, India.
*
Professor & Director, Mangalmay Institute of Engineering & Technology ,
Greater Noida,U.P, India.
Abstract— This paper discusses a white matter lesion (WML)
segmentation scheme for fluid attenuation inversion recovery
(FLAIR) MRI. The method computes the volume of lesions
with subvoxel precision by accounting for the partial volume
averaging (PVA) artifact. As WMLs are associated to stroke
and carotid disease, accurate volume measurements are most
imperative. Manual volume computation is laborious,
subjective, time consuming, and error prone. Automated
methods are a nice substitute since they quantify WML
volumes in an objective, efficient, and trustworthy manner.
PVA is initially modeled with a localized edge strength
measure since PVA resides in the boundaries between tissues.
This map is computed in 3-D and is transformed to a global
representation to raise robustness to noise. Momentous edges
correspond to PVA voxels, which are used to discover the PVA
fraction α (amount of each tissue present in mixture voxels).
Results on simulated and real FLAIR images show high WML
segmentation performance compared to ground truth (98.9%
and 83% overlap, respectively), which outperforms other
process. Lesion load studies are included that automatically
analyze WML volumes for each brain hemisphere separately.
This technique does not necessitate any distributional
assumptions/parameters or training samples and is applied on
a single MR modality, which is a foremost advantage
compared to the traditional methods
grade one tumors under the World Health Organisation's
classification system. In another end, Image retrieval is the
fast growing and challenging research area with regard to
both still and moving images. Many Content Based Image
Retrieval (CBIR) system prototypes have been proposed
and few are used as commercial systems. CBIR aims at
searching image databases for specific images that are
similar to a given query image. It also focuses at evolving
new techniques that support effectual searching and
browsing of large digital image libraries based on
automatically derived imagery features.
Keywords—: Fluid attenuation inversion recovery (FLAIR),
lesion load (LL), MRI brain segmentation, partial volume
averaging (PVA), white matter lesion (WML).
I. INTRODUCTION
To probe in this present scenario, image plays crucial role in
every aspect of business such as business images, satellite
images, medical images and so on. If we analysis these data,
which can unveil useful information to the human users.
But, unfortunately there are certain difficulties to gather
those data in a right way [1]. Hemangioblastomas [1] [2] are
tumors of the central nervous system that originate from the
vascular system usually during middle-age. Occasionally
these tumors crop up in other sites such as the spinal cord
and retina Due to incomplete data, the information gathered
is not processed further for any conclusion.
Hemangioblastomas are most commonly composed of
stromal cells in small blood vessels and usually occur in the
cerebellum, brain stem or spinal cord. They are classed as
ISSN: 2231-5381
Fig 1.1 (A) Affected Brain with tumor. (B) Hemangioblastoma Tumor
identified in axial MR Slice, (C) Manual Annotation of Lesion (D) Final
Segmentation
It is a rapidly expanding research area situated at the
intersection of databases, information retrieval, and
computer vision. Though CBIR is still immature, there has
been abundance of prior work. A cerebral vascular accident,
or stroke, is an acute neurological injury caused by an
http://www.ijettjournal.org
Page 101
International Journal of Engineering Trends and Technology(IJETT) - Volume 4 Issue 1-Jan 2013
interruption of the vascular supply of blood to the brain.
Since blood is a carrier of nutrients and oxygen, the
pretentious neurons begin to die within minutes due to
oxygen and/or nutrient starvation (ischemic stroke) [1].
Stroke can result in significant neurological deficits, leading
to various physical impairments such as sensory motor
paralysis, loss of sensation and motor control, as well as
complexity in interpreting spatial relationships [1]; stroke
can also be fatal. According to the Canadian Heart and
Stroke Society, about 50 000 Canadians endure new or
recurrent strokes each year, which on average means a
stroke occurs every 10 min. It is the third cause of death
behind heart disease and cancer and costs the Canadian
economy roughly $3.6 billion a year in physician services,
hospital costs, lost wages, and decreased productivity [2].
MRI is superior to nonenhanced CT in the detection of
vascular components of the tumor. [8, 10] Contrastenhanced CT has the same sensitivity as nonenhanced MRI;
though, it is inferior to contrast-enhanced MRI. [8]
Contrast-enhanced MRI permits the identification of small
tumor nodules. In addition, MRI is helpful in unraveling
cystic and solid components of the tumor from edema.
Patients with VHL should be screened, and follow-up
studies should be carried out at 6 months. The sensitivity of
MRI increases with the use of gadolinium-based contrast
material. Angiography is better in the detection of small (<
1 cm) vascular tumor components, and it is enhanced for
showing the vascular nature, supply, and drainage of
tumors, compared with CT.[11] Though, CT and MRI
depict tumor cysts better.[12]. Use of a gadolinium-based
contrast agent is mandatory in the evaluation of
hemangioblastomas because it increases the sensitivity for
small, solid lesions. [11, 13] When the diagnosis of
hemangioblastoma is established, a cautious evaluation for
other petite enhancing lesions should be performed because
of the multiple lesions seen in some patients. The presence
of multiple lesions has an important impact on the
prognosis. [6]. In hemangioblastomas located near the pia,
differentiation from meningiomas can be difficult in certain
patients. It is always imperative to look for characteristics
that can aid in diagnosing a vascular tumor, which is
particularly important before surgery. In some patients,
complete removal of the mural nodule is enough; though, in
patients in whom the transition from solid tumor to cystic
tumor with a mural nodule has been observed, the
radiologist should perform careful follow-up imaging to
assess the behavior of the lesions. To reduce the mortality
rates and long-term disabilities associated with stroke,
physicians are investigating magnetic resonance images
(MRI) of the brain to determine if precursors exist.
Identifying early on stages of the disease can lead to new
intervention protocols and therapeutic strategies which
ISSN: 2231-5381
ultimately may reduce the incidence of stroke. MRI are fast
becoming the de facto standard for brain analysis since MR
scanners generate high-resolution images with nonionizing
radiation and they offer a noninvasive view of the softtissue structures of the brain. Through the analysis of a
specific type of MRI, known as fluid attenuation inversion
recovery (FLAIR), researchers have found that abnormal
changes in the white matter, known as white matter lesions
(WML), are a surrogate for future stroke [3]. FLAIR images
have similar tissue contrasts as T2-weighted MRI
(suppressed fat signal, hyper intense in water-based tissues),
except that the cerebrospinal fluid (CSF) signal is nulled for
enhanced discrimination of ischemic pathology [4].
Resultantly, WML appear as hyper intense objects scattered
throughout the white matter in FLAIR MRI. Using the
FLAIR images, lesions were manually outlined by a
specialist and the volumes were found based on the number
of pixels in the region. The volumes of these lesions [lesion
load (LL)] were important markers in determining their
relationship to stroke.
Since there is evidence that relates WML to ischemic
stroke, there is growing interest and research efforts
dedicated to understanding the phenomena of this WML
generation. Discovery of non-genetic and nonage-dependant
factors of WML could aid in the development of new
intervention protocols or therapies (for the mechanism
which causes WML), ultimately reducing a patient’s risk of
stroke.
II. LITERATURE SURVEY
Segmentation Image retrieval is the basic requirement task
in the present scenario. Content Based Image Retrieval is
the popular image retrieval system by which the target
image to be retrieved based on the useful features of the
given image. In other hand, image mining is the arising
concept which can be used to extract potential information
from the general collection of images. Target or close
Images can be retrieved in a little fast if it is clustered in a
right manner. In this paper, the concepts of CBIR and
Image mining have been combined and a new clustering
technique has been introduced in order to increase the speed
of the image retrieval system. Classification followed by
mathematical morphology. Level set evolution with
constant propagation needs to be initialized either
completely inside or outside the tumor and can leak through
weak or missing boundary parts. Replacing the constant
propagation term by a statistical force overcomes these
limitations and results in a convergence to a stable
solutionIn fuzzy classifier systems the classification is
obtained by a number of fuzzy If} Then rules including
linguistic terms such as Low and High that fuzzy each
feature. This paper presents a scheme by which a reduced
linguistic (fuzzy) set of a labelled multi-dimensional data
http://www.ijettjournal.org
Page 102
International Journal of Engineering Trends and Technology(IJETT) - Volume 4 Issue 1-Jan 2013
set can be identified automatically. After the projection of
the original data set onto a fuzzy space, the optimal subset
of fuzzy features is determined using conventional search
techniques. The applicability of this method has been
demonstrated by reducing the number of features used for
the classification of four real-world data sets. This scheme
can also be used to generate an initial rule set for a fuzzy
neural network
III. METHODOLOGY
3.1 Existing Method
Through the analysis of a specific type of MRI, known as
fluid attenuation inversion recovery (FLAIR), researchers
have found that abnormal changes in the white matter,
known as white matter lesions (WML), are a surrogate for
future stroke [3]. FLAIR images have similar tissue
contrasts as T2-weighted MRI (suppressed fat signal, hyper
intense in water-based tissues), except that the cerebrospinal
fluid (CSF) signal is nulled for enhanced discrimination of
ischemic pathology [4]. Resultantly, WML appear as hyper
intense objects scattered throughout the white matter in
FLAIR MRI. Using the FLAIR images, lesions were
manually outlined by a specialist and the volumes were
found based on the number of pixels in the region. The
volumes of these lesions [lesion load (LL)] were important
markers in determining their relationship to stroke. Since
there is evidence that relates WML to ischemic stroke, there
is growing interest and research efforts dedicated to
understanding the phenomena of this WML generation.
Discovery of non-genetic and nonage-dependant factors of
WML could aid in the development of new intervention
protocols or therapies (for the mechanism which causes
WML), ultimately reducing a patient’s risk of stroke. These
medical studies that examine the relationship between
WML and stroke or carotid disease are completed by
humans. Therefore, the number of patients that can be
included in the study is limited (specialists must outline
lesions from several slices per patient, which is a daunting
task). With small sample sizes, it is difficult to study longterm effects or the “true” statistical nature for large patient
cohorts. Furthermore, as with any human-based analysis,
manually acquiring the volumes of the lesions is subjective
(observer dependant), time consuming, laborious, and error
prone. Automated techniques are a great alternative since
they can automatically segment the WML and compute
their volume in a quantitative, efficient, reproducible, and
trustworthy manner. Any numeral of images can be
included in the study and LLs can be obtained within
minutes.
large vessel atherosclerosis and WML [5]. In [6], Moody et
al. examine the relationship between WML and carotid
plaque morphology as an indirect measure of micro embolic
activity. The hypothesize that if micro embolisms from
complicated plaques are a contributing factor to WML, then
these lesions should be more frequent and evident in the
ipsilateral cerebral hemisphere (same side of the body as the
diseased carotid). To test this hypothesis, WML were
manually segmented from FLAIR images of the brain and
the volumes of the lesions from each of the brain’s
hemispheres were computed. Statistical tests found twice as
many WML (higher LL) in the brain hemisphere that is on
the same side of the body (ipsilateral) to complicated
carotid plaques. Similar studies have shown the same results
[5]. Most works for WML segmentation focus on intensity
clustering of multispectral datasets, (T1, T2, PD, FLAIR)
[7]–[11]. In [7] and [8], the authors examine the
clusterability of image classes for intensity-based feature
vectors defined from the coregistered dataset. Although
promising, the need for protracted multiparametric scanning
reduces the appeal of such approaches. Acquisition of
several volumes per patient is expensive, dependent on the
performance of a registration algorithm (which can
introduce errors for over- and under-fitting) and results in
images that are prone to motion artifacts (patients spend
more time in the MR scanner). Moreover, the use of many
images increases memory requirements and the
computational complexity of the algorithm.
3.3 PROPOSED METHOD
There are many works that deal with segmentation of
multiple sclerosis (MS) plaques or lesions in FLAIR MRI.
For example, the works in [12] and [13] use a Bayesian
classifier with the adaptive mixtures method and Markov
random field classifier to automatically estimate and update
the class conditional probability density function and the a
priori probability of each tissue class. MS plaque
segmentations were validated extensively with manual
segmentations and the results are very promising. Although
MS lesions have similar appearance to WML in FLAIR
MRI in many instances, MS lesions can have unique
characteristics that set them apart from WML. Not only the
origin of these diseases is different (MS is an inflammatory
demyelinating disease [14], whereas WML are selective
incomplete white matter infarction [6]), the location of the
lesions can also differ substantially. For example,
juxtacortical lesions (inside the cortex) are specific to MS
and not WML [14]. Consequently, not all MS segmentation
schemes may be applicable for WML detection and
quantification.
3.2 Disadvantages
Obtaining memberships partitioned
The etiology or pathogenesis of WML is still largely
debated. For some time, the predominant notion was that
WML is strongly associated with small vessel disease.
However, as of more recently, there has been an
accumulation of evidence that supports the link between
As will be shown, the parameter α dictates how much of
each tissue is present in a voxel. Consequently, PVA
quantification is usually posed as an estimation problem,
where this variable α is sought. Traditional techniques for
normal T1 or T2 MRI search for a global estimate of this
ISSN: 2231-5381
http://www.ijettjournal.org
Page 103
International Journal of Engineering Trends and Technology(IJETT) - Volume 4 Issue 1-Jan 2013
parameter, i.e., α(y), which quantifies the proportion of one
tissue present in a voxel of intensity y [16], [17].
Unfortunately, previous model-based methods are not
directly applicable since pathology modifies the intensity
distribution in a manner that is difficult to model. Moreover,
neurological MRI often has non-Gaussian or unknown noise
properties, causing techniques that rely on normality to be
inaccurate [18]. To combat these downfalls, this study
focuses on a nonstatistical, image-based PVA modeling
approach for robust segmentation of WML in FLAIR. The
following section details the methods used.
3.4 Algorithm
To prove the efficacy of the proposed method, both
simulated and real images with ground truths are used. To
objectively validate the performance of the proposed
method on these databases, the amount of intersection
between a segmented object and the gold standard is
measured by the dice similarity coefficient (DSC):
(
| ( ) ⋂ ( )|
| ( )| | ( )|
)
( ))
3.5 Flair On Implementation
1)
Simulated Data: A series of FLAIR images with
WML are simulated based on McGill’s Brainweb
database. Brainweb contains a series of T1/T2/PD images
(normal and MS lesions) with ground truth masks for GM,
WM, CSF, and MS lesion classes. Several Brainweb
slices with MS lesions that have similar appearance and
spatial location as WML are used for simulation purposes.
To generate FLAIR images, the WM and GM classes are
joined, resulting in three pure tissue classes: CSF, brain
(GM and WM), and WML. This mask is interpolated by
two to double the size of the mask.
2) Real Data: To validate the algorithm on real data, a total
of 25 images were selected by randomization from the
database. The images were chosen according to uniform
sampling without replacement. The validation database was
given to an expert radiologist (AM) for the manual WML
segmentations (ground truth). AM was blinded to all patient
ISSN: 2231-5381
Advantages
As WML have been shown to be related to carotid disease
ipsilaterally, large differences in the hemispheric LLs could
be an indication of advanced carotid disease. Preliminary
results presented here on the patient data demonstrate how
the proposed works can be used to gather statistics
regarding WML automatically, efficiently, and reliably. The
end goal of these methods is WML segmentation in large
studies to further examine the relationship between WML,
stroke, and carotid disease. Therefore, we are currently
recruiting more patients and preparing to apply this research
on a large patient cohort. However, the methods presented
here give a clear sense of direction and show the potential
of automated analysis for WML segmentation.
IV.WHITE MATTER NEOPLASM DISSECTION
Where A(x) and B(x) are binary masks for the segmentation
and ground truth.
Since the DSC only highlights the amount of overlap
between two sets, to further quantify the efficiencies and
deficiencies of the proposed method, specificity and
sensitivity are used. This line divides the ROC space, where
points above this line represent good classification, whereas
points below this line indicate poor performance. Finally,
the volumes of the lesions are computed with
(∑
information, as well as the segmentation results. The
software used to draw the regions of interest is known as the
Sedeen Viewer.5 It allows users to easily load images, draw
contours for desired regions of interest, as well as manual
editing of the drawn contours.
As will be shown, the parameter α dictates how much of
each tissue is present in a voxel. Consequently, PVA
quantification is usually posed as an estimation problem,
where this variable α is sought. Traditional techniques for
normal T1 or T2 MRI search for a global estimate of this
parameter, i.e., α(y), which quantifies the proportion of one
tissue present in a voxel of intensity y [16], [17].
Unfortunately, previous model-based methods are not
directly applicable since pathology modifies the intensity
distribution in a manner that is difficult to model. Moreover,
neurological MRI often has non-Gaussian or unknown noise
properties, causing techniques that rely on normality to be
inaccurate [18]. To combat these downfalls, this study
focuses on a nonstatistical, image-based PVA modelling
approach for robust segmentation of WML in FLAIR. The
following section details the methods used
List of Modules
1.
PVA Model
2.
Edge-Based PVA Modeling
3.
Fuzzy Edge Model
4.
Global Edge Description
5.
Estimating α
http://www.ijettjournal.org
Page 104
International Journal of Engineering Trends and Technology(IJETT) - Volume 4 Issue 1-Jan 2013
PVA Model
compute this metric, the traditional magnitude of the
gradient, i.e., g, is first estimated by
PVA generates an image intensity which is linearly
dependant on the proportion of each tissue in the voxel. In
neuro MRI, where two tissue types mix per PVA voxel
[17], the intensities of these mixels,1 Yjk (x), are
determined by the proportion of the first tissue j, in
comparison to that of the second tissue k, as in
( )
( )
( )
(
( ))
( )
where Yj (x) is the intensity value drawn from first tissue’s
intensity distribution pj (y) at spatial location x = (x1, x2 ) ∈
Z2 , Yk (x) is the intensity of the second tissue ∼pk (y),
andα ∈ [0, 1] is the proportion of tissue j present at x (the
remainder of the voxel is a fraction of tissue k, i.e., 1 − α).
Using this mathematical relationship that describes PVA in
terms of the intensities of mixtures voxels, we will quantify
PVA in a new way based on the edge content of the image.
Edge-Based PVA Modeling
To examine the edge content in the PVA regions, consider
the ideal signal model. Since edge content is described by
the gradient, the gradient of these equations are taken
resulting in
(
)
(
)
Where α is the change in the proportion of tissues parameter
(dictates how the proportion of one tissue changes as a
function of space). Solving for the change in the proportion
of tissues that result in two PVA quantifiers.
Each PVA measure α jk is a normalized, class-specific
representation of edge information in PVA regions. It is a
normalized representation because the largest possible value
of Y jk is Ij − Ik (maximum intensity change in one pixel
step) and the minimum is 0 in a constant region, resulting in
0 ≤ αjk ≤ 1. Since these class-specific variables describe
PVA in terms of the gradient, this study focuses on an edgebased estimate for αand uses it to decode the proportion of
tissues parameter α.
Fuzzy Edge Model
To estimate α(x), a fuzzy technique based on the cumulative
distribution function (CDF) of the gradient is employed. To
ISSN: 2231-5381
‖
√|
‖
|
|
|
Where the Sobel operator is used. The probability
distribution function (PDF) of the gradient pG (g) is
computed next, and based on this PDF, the CDF of the
gradient magnitude is found and used as an estimate for the
edge information in the image.
( )
(
)
∑
()
Where α(g) ∈ [0, 1]. This nonlinear fuzzification of the
edge information quantifies the “certainty of edge
presence.” Note that this parameter is expressed as a
function of the gradient and to be used to approximate α(x),
α(g) is mapped back to the spatial domain: α(g) → α(x).
This fuzzy edge measure assigns large and similar values to
significant edges, despite them occurring over a wide range
of g and with few occurrences. It groups significant edges,
while suppressing the irrelevant ones. As this edge measure
is normalized, i.e., 0 ≤ α(x) ≤ 1, and representative of the
edge information in the image, it is used to represent PVA
in the image.
Global Edge Description
PVA occurs over specific intensity ranges, and moreover,
with high edge values. These two features, intensity and
edge strength, are coupled together in the following section
to arrive at a new and denoized version of the fuzzy edge
metric. Initially, edge and intensity information is coupled
through the conditional PDF of α(x), for a particular
intensity y, by
( )|
( ( )
|
)
( )
|
Where 0 ≤ a ≤ 1, 0 ≤ y ≤ ymax, a is the realization of α(x),
and y max is the maximum gray level in the image. This
PDF quantifies the distribution of the edge information α(x)
for a specific gray level y. Because it was computed on the
entire image (or volume for 3-D approaches), it describes
the global clustering trend of edge information as a function
of intensity. Generally, in flat regions (pure tissues) there is
clustering in the PDF for low edge values at corresponding
intensities. Across anatomical boundaries (PVA), high edge.
http://www.ijettjournal.org
Page 105
International Journal of Engineering Trends and Technology(IJETT) - Volume 4 Issue 1-Jan 2013
Estimation of the PDF this way automatically classifies the
voxel y as belonging to either the pure tissue or PVA class.
Pure regions correspond to maxima in p (a = 0|y) and
minima in p (a = 1|y), whereas minima in p (a = 0|y) and
maxima in p (a = 1|y) indicate with high likelihood that
these voxels belong to a PVA region. To determine the
global estimate of α(y), the conditional expectation operator
is used. It offers the best prediction of α given that the
intensity is y in the mean squared error sense. The result is
an enhanced edge map α(y),
( )
( )|
∑
( )|
( )|
(
(
| )
( | )
( )|
(
| )
| )
Which provides a global representation of the edge
information in the image indicating that the quantification
of PVA content is directly proportional to the probability
that a voxel is located on an edge.
Estimating α
To decode α(y), regions of α(y) are retained, while others
discarded. Recall that the maxima of p (a = 1|y) dictate
which voxels y are most likely PVA (maximally edgy),
while the minima are correlated with voxels y from pure
tissue classes (minimally edgy or flat). Ideally, in flat
regions (pure tissues), there should be no edge information,
but noise generates “artificial” edginess, causing the
minima of α(y) to be nonzero in these regions. To account
for the relative nature of α(y), an adaptive threshold is
applied.
Where tL and tR are the left and right thresholds,
respectively, and mink is the minimum of α(y)
corresponding to tissue k, minj is the minimum of α(y)
corresponding to pure tissue j, and maxjk is the maximum
of the PVA pulse that describes the mixture of tissues j and
k. The minima and maxima values are easily found with a
peak-finding algorithm, which uses derivative information
to find optima.
V. EXPERIMENT RESULT
The programming language used with MATLAB is usually
referred to as MATLAB script or M-script. After becoming
familiar with the basic syntax of the M-script, a number of
useful utilities are available to you that allow you to make
extended uses of MATLAB. You can, for example, write
programs that involve simulation. You can also create
graphics, web pages, and GUI applications. When you
develop programs using MATLAB, you can output the
results to a number of media, including graphics files,
HTML pages, PDF files, and Word documents. You can
also connect up MATLAB with other applications, such as
Excel or Lab View to make extended uses of it. Since it is
programmed in part using Java, you can modify it in the
background using Java. The feasibility of the project is
analyzed in this phase and business proposal is put forth
with a very general plan for the project and some cost
estimates. During system analysis the feasibility study of
the proposed system is to be carried out. This is to make
sure that the proposed system is not a burden to the
company. For feasibility analysis, some understanding of
the major requirements for the system is essential.
SCREENSHOTS
INPUT IMAGE
An adaptive threshold that retains voxels most likely (in a
probabilistic sense) to contain mixture components are
computed for the left and right side of each PVA pulse
ISSN: 2231-5381
http://www.ijettjournal.org
Page 106
International Journal of Engineering Trends and Technology(IJETT) - Volume 4 Issue 1-Jan 2013
Fig: 5.3 Manual Annotations
INITIAL ITERATION
Fig: 5.1 Input MR Image
7.2 SEGMENTED IMAGE BASED ON ALGORITHM
PROCESSING IMAGE FIGURE 1
Fig: 5.4 Computing the points
FINAL ITERATION
Fig: 5.2 Point Analysis
PROCESSING STAGE
Fig: 5.5 Segmenting the tumor region
ISSN: 2231-5381
http://www.ijettjournal.org
Page 107
International Journal of Engineering Trends and Technology(IJETT) - Volume 4 Issue 1-Jan 2013
TUMOR REGION EXTRACTION
PROCESSING
Fig: 5.6: Segmented tumor region
Fig: 5.9 Point Analysis
INITIAL ITERATION
Fig: 5.7 Segmented tumors in surf analysis
7.3 SEGMENTED IMAGE BASED ON ALGORITHM
PROCESSING IMAGE FIGURE 2
Fig: 5.10 Manual Annotation
FINAL ITERATION
Fig: 5.8 Initial Stage Of Another Image
Fig: 5.11 Computing the points
ISSN: 2231-5381
http://www.ijettjournal.org
Page 108
International Journal of Engineering Trends and Technology(IJETT) - Volume 4 Issue 1-Jan 2013
CELL REGION EXTRACTION
hemispheres of the brain separately, which has the potential
to advance medical research on stroke and carotid disease.
6.2 FUTURE WORKS
Preliminary results presented here on the patient data
demonstrate how the proposed works can be used to gather
statistics regarding WML automatically, efficiently, and
reliably. The end goal of these methods is WML
segmentation in large studies to further examine the
relationship between WML, stroke, and carotid disease. It is
most important that large databases are included to gain true
knowledge of the underlying phenomena, and therefore, we
are currently recruiting more patients and preparing to apply
this research on a large patient cohort. Though, the methods
presented here give a clear sense of direction and show the
potential of automated analysis for WML segmentation.
Fig: 5.12 Segmented cell region
SURF ANALYSIS OF THE PROCESSED IMAGE
REFERENCE
[1] F. H. Martini, Fundamentals of Anatomy and Physiology, 5th ed.
Fig: 5.13 Segmented cell region in surf
VI. CONCLUSION
6.1 CONCLUSION
This paper proposes a novel PVA quantification scheme
that not only robustly segments WML in FLAIR MRI, but
the other tissue classes as well. It focuses on a global edgebased approach since PVA voxels reside in boundaries
between tissues. Results for simulated images show
excellent results (WML, brain, and CSF are segmented with
an accuracy of 98.5%, 99.9%, and 98.9%, respectively).
Results on real images further show the utility of the work,
as the WML segmentations show high correlation with the
expert segmentations (average overlap of 83%). LL studies
on a patient database show that the technique can be used to
measure the volume of the WML in the left and right
ISSN: 2231-5381
Englewood Cliffs, NJ: Prentice-Hall, 2001.
[2] Canadian Hearth and Stroke Association. (2011). “Stroke statistics,”
World
Wide
Web,
[Online].
Available:
http://www.heartandstroke.com
[3] A. J. Fox, V. C. Hachinski, H. J. M. Barnett, J. Y. Streifler, M.
Eliasziw, O. R. Benavente, and S. Alamowitch, “Prognostic
importance of leukoaraiosis in patients with symptomatic internal
carotid artery stenosis,” Stroke, vol. 33, pp. 1651–1655, 2002.
[4] P. Malloy, S. Correia, G. Stebbins, and D. H. Laidlaw, “Neuroimaging
of white matter in aging and dementia,” Clin. Neuropsych., vol. 21,
pp. 73– 109, 2007.
[5] R. E. Murphy, A. R. Moody, P. S. Morgan, and A. L. Martel, “Brain
white matter hyperintensities are associated with carotid intraplaque
hemorrhage,” Radiology, vol. 31, no. 1, pp. 202–209, 2007.
[6] N. Altaf, L. Daniels, P. Morgan, J. Lowe, J. Gladman, S. MacSweeney,
A. Moody, and D. Auer, “Cerebral white matter hyper intense lesions
are associated with unstable carotid plaques,” Eur. J. Vasc. and
Endovasc. Surg., vol. 31, pp. 8–13, 2006.
[7] Z. Lao, D. Shen, A. Jawad, B. Karacali, D. Liu, E. Melhem, R. Bryan,
and C. Davatzikos, “Automated segmentation of white matter lesions
in 3D brain MRI, using multivariate pattern classification,” in Proc.
IEEE Int. Symp. Biomed. Imaging (ISBI), 2006, pp. 307–310.
[8] P. Anbeek, K. L. Vincken,M. J. van Osch, R. H. Bisschops, and J. van
der Grond, “Automatic segmentation of different-sized white matter
lesions by voxel probability estimation,” Med. Image Anal., vol. 8,
no. 3, pp. 205– 215, 2004.
[9] P. Anbeek, K. Vincken, M. van Osch, R. Bisschops, and J. van der
Grond, “Probabilistic segmentation of white matter lesions in MR
imaging,” NeuroImage, vol. 21, no. 3, pp. 1037–44, 2004.
[10] R. de Boer, F. van der Lijn, H. A. Vrooman, M.W. Vernooij, M. A.
Ikram, M. M. Breteler, and W. J. Niessen, “Automatic segmentation
of brain tissue and white matter lesions in MRI,” in Proc. IEEE Int.
Symp. Biomed. Imag., 2007, pp. 652–655.
[11] R. de Boer, H. A. Vrooman, F. van der Lijn,M.W. Vernooij, M. A.
Ikram, A. van der Lugt, M. M. Breteler, and W. J. Niessen, “White
matter lesion extension to automatic brain tissue segmentation on
MRI,” Neuroimage,vol. 45, no. 4, pp. 1151–1161, 2009.
[12] R. Khayatia, M. Vafadusta, F. Towhidkhaha, and S.M. Nabavib,
“Fully automatic segmentation of multiple sclerosis lesions in brain
MR flair images using adaptive mixtures method and Markov
random field model,” Comput. Biol. Med., vol. 38, no. 3, pp. 379–
390, 2008.
[13] R. Khayatia, M. Vafadusta, F. Towhidkhaha, and S. M. Nabavib, “A
novel method for automatic determination of different stages of
multiple sclerosis lesions in brain MR flair images,” Comput. Med.
Imag. Graph., vol. 32,no. 2, pp. 124–133, 2008.
http://www.ijettjournal.org
Page 109
International Journal of Engineering Trends and Technology(IJETT) - Volume 4 Issue 1-Jan 2013
[14] F. Barkhof and R. Smithuis. (2007). “Multiple sclerosis,” Radiology
Assistant,
[Online].
Available:
http://www.radiologyassistant.nl/en/4556dea65db62
[15] C. R. Jack, P. C. O’Brien, D. W. Rettman, M. M. Shiung, Y. C. Xu, R.
Muthupillai, A. Manduca, R.Avula, and B. J. Erickson, “Flair
histogram segmentation for measurement of leukoaraiosis volume,”
J. Mag. Res. Img., vol. 14, no. 6, pp. 668–676, 2001.
[16] K. V. Leemput, F. Maes, D. Vandermeulen, and P. Suetens, “A
unifying framework for partial volume segmentation of brain MR
images,” IEEE Trans. Med. Imag., vol. 22, no. 1, pp. 105–119, Jan.
2003.
[17] M. A. Gonz´alez Ballester, A. P. Zisserman, and M. Brady,
“Estimation of the partial volume effect in MRI,” Med. Image Anal.,
vol. 6, no. 4, pp. 389–405, 2002.
[18] A. Khademi, D. Hosseinzadeh, A. Venetsanopoulos, and A. R.
Moody, “Nonparametric statistical tests for exploration of correlation
and nonstationarity in images,” in Proc. Int. Conf. Digital Signal
Process., 2009, pp. 1–6.
[19] A. Khademi, A. Venetsanopoulos, and A. R. Moody, “Automatic
contrast enhancement of white matter lesions in FLAIR MRI,” in
Proc. IEEE Int. Symp. Biomed. Imag., 2009, pp. 1–4.
[20] P. Meer and B. Georgescu, “Edge detection with embedded
confidence,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 23, no. 12,
pp. 1351–1365, Dec. 2001.
[21] A. Khademi, A. Venetsanopoulos, and A. Moody, “Edge-based PVA
estimation in FLAIR MRI with WML,” in Proc. IEEE Eng. Med.
Biol. Soc. Conf., 2010, pp. 6114–6117.
[22] P. H. Kvam and B. Vidakovic, Nonparametric Statistics with
Applications to Science and Engineering. New York: WileyInterscience, 2007.
[23] A. Popovic, M. de la Fuente, M. Engelhardt, and K. Radermacher,
“Statistical validation metric for accuracy assessment in medical
image segmentation,” Int. J. Comput. Assist. Radiol. Surg., vol. 2, pp.
169–181, 2007.
[24] C. Lisanti, C. Carlin, K. P. Banks, and D.Wang, “NormalMRI
appearance and motion-related phenomena of CSF,” Amer. J.
Roentgenol., vol. 188, no. 3, pp. 716–725, 2007.
[25] Vijayakumar, B., and Ashish Chaturvedi. "Automatic Brain Tumors
Segmentation of MR Images using Fluid Vector Flow and Support
Vector Machine." Research Journal of Information Technology 4.
[26] Vijayakumar, B., and Ashish Chaturvedi. "Tumor Cut-Segmentation
and Classification of MR Images using Texture Features and Feed
Forward Neural Networks." European Journal of Scientific Research
85.3 (2012): 363-372
[27] Vijayakumar, B., and Ashish Chaturvedi, “Idiosyncrasy Dissection
and Assorting of Brain MR Images Instigating Kernel Corroborated
Support Vector Machine,” Archive Des Science
ISSN: 2231-5381
http://www.ijettjournal.org
Page 110