More Related Content
Similar to 34 107-1-pb (20)
More from Mahendra Sisodia
More from Mahendra Sisodia (11)
34 107-1-pb
- 1. ISSN: 2277 – 9043
International Journal of Advanced Research in Computer Science and Electronics Engineering
Volume 1, Issue 2, April 2012
Brain Segmentation using Fuzzy C means clustering
to detect tumour Region.
Prof. A.S.Bhide1 Priyanka Patil2, Shraddha Dhande3
1 2
Electronics and Communication Engineering, Electronics and Communication Engineering,
North Maharashtra University, Jalgaon, India. North Maharashtra University, Jalgaon, India
3
Electronics and Communication Engineering,
Vishwakarma Institute of Technology, Pune, India
Abstract There are two classifications exist to recognize a pattern, and
Tumor segmentation from MRI data is an they are supervised classification and unsupervised
important but time consuming manual task performed by classification. A commonly used unsupervised classification
medical experts. The research which addresses the method is a Fuzzy C Means algorithm [2].
diseases of the brain in the field of the vision by computer
is one of the challenges in recent times in medicine, the
Clustering is a process of partitioning or grouping a given
engineers and researchers recently launched challenges to
carry out innovations of technology pointed in imagery. sector unlabeled pattern into a number of clusters such that
This paper focuses on a new algorithm for brain similar patterns are assigned to a group, which is considered
segmentation of MRI images by fuzzy c means algorithm to as a cluster. There are two main approaches to clustering
diagnose accurately the region of cancer. In the first step it which is crisp clustering and fuzzy clustering techniques .One
proceeds by noise filtering later applying FCM algorithm to of the characteristic of crisp clustering method is that the
segment only tumor area .In this research multiple MRI images boundary between clusters is fully defined but in many real
of brain can be applied detection of glioma (Tumor) growth by cases the boundary between clusters cannot be clearly defined.
advanced diameter technique. Some patterns may belong to more than one cluster. In such
cases, the fuzzy clustering method provides a better and more
Index Terms - Brain tumor, MRI, Imaging, Segmentation.
useful method to classify these patterns. Fuzzy clustering
method and its derivatives have been used for pattern
1. INTRODUCTION
recognition, classification, data mining, and image
segmentation It has also been used for medical image data
Brain tumor is an abnormal mass of tissue in which cells
analysis and modelling etc. Clustering is used for pattern
grow and multiply uncontrollably, seemingly unchecked by
recognition in image processing, and usually requires a high
the mechanisms that control normal cells. Brain tumours can
volume of computation. This high volume computation
be primary or metastatic, and either malignant or benign. A
requires considerable amount of memory which may lead to
metastatic brain tumor is a cancer that has spread from
frequent disk access, making the process inefficient. With the
elsewhere in the body to the brain [3].
development of affordable high performance parallel systems,
Magnetic Resonance Imaging (MRI) is an advanced medical parallel algorithms may be utilized to improve performance
imaging technique used to produce high quality images of the and efficiency of such tasks. The computation speed and
parts contained in the human body MRI imaging is often used memory requirement needed for executing FCM is a big
when treating brain tumours, ankle, and foot. From these hurdle which tried to overcome in this report. In FCM, the
high-resolution images, we can derive detailed anatomical cluster centre initialized by random numbers and it requires
information to examine human brain development and more number of iteration for converging to a final actual
discover abnormalities. Nowadays there are several cluster centre [2].
methodology for classifying MR images, which are fuzzy
The tumor volume is prognostic factor in the treatment of
methods, neural networks, atlas methods, knowledge based
malignant tumours. Manual segmentation of brain tumours
techniques, shape methods, variation segmentation. Image
from MR images is a challenging and time consuming task. In
segmentation is the primary step in image analysis, which is
this study a new approach has been discussed to detect the
used to separate the input image into meaningful regions.
85
All Rights Reserved © 2012 IJARCSEE
- 2. ISSN: 2277 – 9043
International Journal of Advanced Research in Computer Science and Electronics Engineering
Volume 1, Issue 2, April 2012
volume of brain tumor using diameter and graph based image sharpening is to make the tumor edges, contour lines
method to find the volume. Here MRI data set from patients and image details clearer. Same process will be applied to the
were collected. The graph based on pixel value is drawn real target image.
taking the various points from the tumor cells lies in the
original position from the affected region. Here the affected
region is considered as ellipse shape and the volumes have
3.1 Database Collection
been calculated from it. In this system the mean has been
found from the volumes grown in the affected region. The 3.1.1 CT Scan
experimental results show that 96% brain tumor growth and CT scans are a specialized type of x-ray. The
volume can be measured by graph and diameter method [6]. patient lies down on a couch which slides into a large circular
opening. The x-ray tube rotates around the patient and a
2. LITERATURE SURVEY computer collects the results. These results are translated into
The previous methods for brain tumor segmentation are images that look like a "slice" of the person.
thresholding, region growing & clustering.
Thresholding is the simplest method of image segmentation. Sometimes a radiologist will decide that contrast agents
From a greyscale image, thresholding can be used to should be used. Contrast agents are iodine based and are
create binary images. During the thresholding process, absorbed by abnormal tissues. They make it easier for the
individual pixels in an image are marked as "object" pixels if doctor to see tumors within the brain tissue. There are some
their value is greater than some threshold value (assuming an (rare) risks associated with contrast agents and you should
object to be brighter than the background) and as make sure that you discuss this with the doctor before arriving
"background" pixels otherwise. This convention is known for the examination.
as threshold above. Variants include threshold below, which is
opposite of threshold above; threshold inside, where a pixel is CT is very good for imaging bone structures. In fact, it's
labelled "object" if its value is between two thresholds; and usually the imaging mode of choice when looking at the inner
threshold outside, which is the opposite of threshold inside. ears. It can easily detect tumours within the auditory canals
Typically, an object pixel is given a value of “1” while a and can demonstrate the entire cochlea on most patients.
background pixel is given a value of “0.” Finally, a binary
image is created by colouring each pixel white or black,
depending on a pixel's labels. 3.2.2 MRI
The major drawback to threshold-based approaches is that MRI is a completely different. Unlike CT it
they often lack the sensitivity and specificity needed for uses magnets and radio waves to create the images. No x-rays
accurate classification. are used in an MRI scanner. The patient lies on a couch that
The first step in region growing is to select a set of seed looks very similar the ones used for CT. They are then placed
points. Seed point selection is based on some user criterion in a very long cylinder and asked to remain perfectly
(for example, pixels in a certain gray-level range, pixels still. The machine will produce a lot of noise and
evenly spaced on a grid, etc.). The initial region begins as the examinations typically run about 30 minutes.
exact location of these seeds. The cylinder that you are lying in is actually a very large
The regions are then grown from these seed points to adjacent magnet. The computer will send radio waves through your
points depending on a region membership criterion. The body and collect the signal that is emitted from the hydrogen
criterion could be, for example, pixel intensity, gray level atoms in your cells. This information is collected by an
texture or colour. antenna and fed into a sophisticated computer that produces
Since the regions are grown on the basis of the criterion, the the images. These images look similar to a CAT scan but they
image information itself is important. For example, if the have much higher detail in the soft tissues. Unfortunately,
criterion were a pixel intensity threshold value, knowledge of MRI does not do a very good job with bones.
the histogram of the image would be of use, as one could use One of the great advantages of MRI is the ability to change
it to determine a suitable threshold value for the region the contrast of the images. Small changes in the radio waves
membership criterion and the magnetic fields can completely change the contrast of
the image. Different contrast settings will highlight different
types of tissue.
3. BRAIN MRI IMAGE PREPROCESSING
Another advantage of MRI is the ability to change the imaging
plane without moving the patient. If you look at the images to
In order to improve the visual effects of the image
the left you should notice that they look very different. The
for further image recognition, MRI image pre-processing is top two images are what we call axial images. This is what
needed, mainly including colour image greyscale, image you would see if you cut the patient in half and looked at them
smoothing and sharpening and so on. Image smoothing is to from the top. The image on the bottom is a coronal
eliminate noise and improve image quality. The purpose of image. This slices the patient from front to back. Most MRI
86
All Rights Reserved © 2012 IJARCSEE
- 3. ISSN: 2277 – 9043
International Journal of Advanced Research in Computer Science and Electronics Engineering
Volume 1, Issue 2, April 2012
machines can produce images in any plane. CT can not do sensor accuracy typically 10 or 12 bits per sample and to
this. guard against round off errors in computations. Sixteen bits
Contrast agents are also used in MRI but they are not made of per sample (65,536 levels) is a convenient choice for such
iodine. There are fewer documented cases of reactions to uses, as computers manage 16-bit words efficiently. The TIFF
MRI contrast and it is considered to be safer than x-ray and the PNG among other image file formats supports 16-bit
dye. Once again, you should discuss contrast agents with your grayscale natively, although browsers and many imaging
physician before you arrive for the examination. programs tend to ignore the low order 8 bits of each pixel.
No matter what pixel depth is used, the binary
3.2 Grayscale representations assume that 0 is black and the maximum value
In photography and computing, a grayscale or 255 at 8 bpp, 65,535 at 16 bpp, etc. is white, if not otherwise
grayscale digital image is an image in which the value of each noted [1].
pixel is a single sample, that is, it carries only intensity
information. Images of this sort, also known as black-and- 3.3 Converting Color to Grayscale
white, are composed exclusively of shades of gray, varying Conversion of a color image to grayscale is not
from black at the weakest intensity to white at the strongest. unique; different weighting of the color channels effectively
Grayscale images are distinct from one-bit black- represents the effect of shooting black-and-white film with
and-white images, which in the context of computer imaging different-colored photographic filters on the cameras. A
are images with only the two colors, black, and white also common strategy is to match the luminance of the grayscale
called bi-level or binary images. Grayscale images have many image to the luminance of the color image [1].
shades of gray in between. Grayscale images are also called To convert any color to a grayscale representation of
monochromatic, denoting the absence of any chromatic its luminance, first one must obtain the values of its red,
variation. green, and blue (RGB) primaries in linear intensity encoding,
Grayscale images are often the result of measuring by gamma expansion. Then, add together 30% of the red
the intensity of light at each pixel in a single band of the value, 59% of the green value, and 11% of the blue value
electromagnetic spectrum e.g. infrared, visible light, these weights depend on the exact choice of the RGB
ultraviolet, etc, and in such cases they are monochromatic primaries, but are typical. Regardless of the scale employed
proper when only a given frequency is captured. But also they 0.0 to 1.0, 0 to 255, 0% to 100%, etc., the resultant number is
can be synthesized from a full color image. the desired linear luminance value; it typically needs to be
The intensity of a pixel is expressed within a given gamma compressed to get back to a conventional grayscale
range between a minimum and a maximum, inclusive. This representation [1].
range is represented in an abstract way as a range from 0 This is not the method used to obtain the luma in the
means total absence, black and 1 means total presence, white Y'UV and related color models, used in standard color TV and
with any fractional values in between. video systems as PAL and NTSC, as well as in the L*a*b
Another convention is to employ percentages, so the color model. These systems directly compute a gamma-
scale is then from 0% to 100%. This is used for a more compressed luma as a linear combination of gamma-
intuitive approach, but if only integer values are used, the compressed primary intensities, rather than use linearization
range encompasses a total of only 101 intensities, which are via gamma expansion and compression [1].
insufficient to represent a broad gradient of grays. Also, the To convert a gray intensity value to RGB, simply set
percentile notation is used in printing to denote how much ink all the three primary color components red, green and blue to
is employed in half toning, but then the scale is reversed, the gray value, correcting to a different gamma if necessary.
being 0% the paper white or no ink and 100% a solid black or
full ink. 3.4 Filtering an Image
In computing, although the grayscale can be Image filtering is useful for many applications,
computed through rational numbers, image pixels are stored in including smoothing, sharpening, removing noise, and edge
binary, quantized form. Some early grayscale monitors can detection. A filter is defined by a kernel, which is a small
only show up to sixteen (4-bit) different shades, but today array applied to each pixel and its neighbors within an image.
grayscale images as photographs intended for visual display In most applications, the center of the kernel is aligned with
both on screen and printed are commonly stored with 8 bits the current pixel, and is a square with an odd number 3, 5, 7,
per sampled pixel, which allows 256 different intensities i.e., etc. of elements in each dimension. The process used to apply
shades of gray to be recorded, typically on a non-linear scale. filters to an image is known as convolution, and may be
The precision provided by this format is barely sufficient to applied in either the spatial or frequency domain [1].
avoid visible banding artifacts, but very convenient for Within the spatial domain, the first part of the
programming due to the fact that a single pixel then occupies convolution process multiplies the elements of the kernel by
a single byte. the matching pixel values when the kernel is centered over a
Technical uses in medical imaging or remote sensing pixel. The elements of the resulting array which is the same
applications often require more levels, to make full use of the size as the kernel are averaged, and the original pixel value is
87
All Rights Reserved © 2012 IJARCSEE
- 4. ISSN: 2277 – 9043
International Journal of Advanced Research in Computer Science and Electronics Engineering
Volume 1, Issue 2, April 2012
replaced with this result. The CONVOL function performs centroid or prototypes. Many techniques have been developed
this convolution process for an entire image. Within the for clustering data. In this report c-means clustering is used.
frequency domain, convolution can be performed by It’s a simple unsupervised learning method which can be used
multiplying the FFT of the image by the FFT of the kernel, for data grouping or classification when the number of the
and then transforming back into the spatial domain. The clusters is known. It consists of the following steps.
kernel is padded with zero values to enlarge it to the same size
as the image before the forward FFT is applied. These types of Step 1:
filters are usually specified within the frequency domain and
do not need to be transformed. IDL's DIST and HANNING Choose the number of clusters - K
functions are examples of filters already transformed into the
frequency domain [1]. Step 2:
Since filters are the building blocks of many image
Set initial centers of clusters c1, c2… ck;
processing methods, these examples merely show how to
apply filters, as opposed to showing how a specific filter may Step 3:
be used to enhance a specific image or extract a specific
shape. This basic introduction provides the information Classify each vector
necessary to accomplish more advanced image-specific
processing. filters can be used to compute the first derivatives x [x , x ,....x ] T into the closest centre ci by
of an image [1].
Euclidean distance measure
3.5 Median filter
In image processing, it is often desirable to be able to ||xi-ci ||=min || xi -ci||
perform some kind of noise reduction on an image or signal.
The median filter is a nonlinear digital filtering technique, Step 4:
often used to remove noise. Such noise reduction is a typical
pre-processing step to improve the results of later processing Recomputed the estimates for the cluster centers ci
(for example, edge detection on an image). Median filtering is Let ci = [ci1 ,ci2 ,....cin ] T
very widely used in digital image processing because, under
certain conditions, it preserves edges while removing noise cim be computed by,
cim = ∑xli ∈ Cluster(Ixlim)
4. FUZZY C MEANS ALGORITHM
Ni
The goal of a clustering analysis is to divide a given set of
data or objects into a cluster, which represents subsets or a Where, Ni is the number of vectors in the i-th cluster.
group. The partition should have two properties:
Step 5:
1. Homogeneity inside clusters: the data, which belongs to one
cluster, should be as similar as possible. If none of the cluster centers (ci =1, 2,…, k) changes in step 4
stop; otherwise go to step 3.
2. Heterogeneity between the clusters: the data, which
belongs to different clusters, should be as different as 5. IMPLEMENTATION OF FCM
possible.
The original MRI image of brain is as follows
The membership functions do not reflect the actual data
distribution in the input and the output spaces. They may not
be suitable for fuzzy pattern recognition. To build
membership functions from the data available, a clustering
technique may be used to partition the data, and then produce
membership functions from the resulting clustering.
Clustering is a process to obtain a partition P of a set E of N
objects Xi (i=1, 2,…, N), using the resemblance or
dissemblance measure, such as a distance measure d. A
partition P is a set of disjoint subsets of E and the element Ps
of P is called cluster and the centers of the clusters are called
88
All Rights Reserved © 2012 IJARCSEE
- 5. ISSN: 2277 – 9043
International Journal of Advanced Research in Computer Science and Electronics Engineering
Volume 1, Issue 2, April 2012
structures such as white and gray matter as well as their large
variability.
The filtered image is converted to grey in preprocessing and
FCM is applied to it gives the segmented tumor.
Fig 1.origional MRI
Original MRI image is filtered by using median filter. The
filtered image is shown below.
Fig.3 segmented tumor.
6. DIAMETER METHOD FOR VOLUME
CALCULATION
Diameters of tumor were manually measured on MRI films
with callipers. In each case where a second lesion was present,
or the shape of the lesion was best characterized by two
ellipsoids, a second set of three diameters was also recorded,
and the volumes were summed. If one or more necrotic or
cystic areas were thought to be present, additional diameters
for the cystic component were recorded, and the computed
cystic volume was Subtracted from the overall volume
Readers were not aware that an end point of this study was the
determination of how many sets of diameters (one v two v
Fig.2 filtered image three diameter measurements) were thought to be required to
accurately characterize the lesion volume [6]. The formula
The algorithm of fuzzy c-means (fuzzy c-means) is a used to compute volumes was the standard volume of an
classification algorithm based on fuzzy optimization of a ellipsoid, as follows:
quadratic criterion of classification where each class is
represented by its center of gravity.
The algorithm requires knowing the number of classes in V = 4/3 pi (a *b*c*) …….. (1)
advance and generates classes through an iterative process
minimizing an objective function. Thus, it provides a fuzzy Where a, b, and c are the three radii (half the diameters). In
partition of the image by giving each pixel a degree of addition to the total volume, the individual diameters were
belonging to a given region. also recorded to allow analysis on a single- or dual-diameter
Segmentation of anatomical structures is a critical task in basis, ie, diameter or area rather than a volume estimate.
medical image processing, with a large range of applications
going from visualization to diagnosis.
For example, to delineate structures in the mid-sagittal plane
of the brain in the context of a pre-operative planning, an
accurate segmentation of the hemispheres, and especially of
their internal faces, is needed. In such a task, the main
difficulties are the non-homogeneous intensities within the
same class of tissue, and the high complexity of anatomical
89
All Rights Reserved © 2012 IJARCSEE
- 6. ISSN: 2277 – 9043
International Journal of Advanced Research in Computer Science and Electronics Engineering
Volume 1, Issue 2, April 2012
on a criterion introduced for fuzzy regions. The main
advantage of this method is that, first it requires no prior
information on the images to segment. The computation speed
is fast so less execution time. Less memory requirement.MRI
images are highly weighted so other methods consume high
iteration and also more sensitive to the initialization of
cluster centres but using FCM less iterations are needed in
clustering
Fig.4.Diameter Method Tumor volume is an important diagnostic indicator in
treatment planning and results assessment for brain tumor.
In this diameter method the volume has been calculated using The measurement of brain tumor volume can assist tumor
formula 1. Volume calculation has been done in a step by step staging for brain tumor volume measurements is developed
process from our input MRI image; the affected region has which overcome the problem of inter-operator variance,
been changed as ellipse or circle shape. In formula 1 the besides partial volume effects and shows satisfactory
volume has been found from MRI data set using different performance for segmentation. This method is applied to 8-
parameters (a, b, c). to find out the volume from MRI data set- tumor contained MRI slices from 2 brain tumor patients_ data
1. This result will produce out the volume of brain tumor. sets of different tumor type and shape, and better
Again this same diameter method has been used and has found segmentation results are achieved. In this paper a new
out the volume -2 up to N times. N is a number of steps to approach has been discussed to detect the volume of brain
determine the volume by different parameters as shown in fig. tumor using diameter and graph based method to find the
The mean has been measured from these volumes using volume.
formula-3 which is equal to average of volumes calculated
using different parameters. From this mean the volume of References
glioma has been determined from day by day MRI report. The [1] W. Gonzalez, “Digital Image Processing”, 2nd ed. Prentice
graph shows the brain tumor growth affected in brain cells [6]. Hall, Year of Publication 2008, Page no 378.
Suppose if we have 5 MRI images of brain if these MRI are [2] S. Murugavalli, V. Rajamani, “A high speed parallel fuzzy
applied to FCM algorithm then it will segment all the tumor c-mean algorithm for brain tumour segmentation”,” BIME
region and calculates the area and volume of individual MRI. Journal”, Vol. no: 06, Issue (1), Dec., 2006
If the tumor is growing day by day then it will plot then tumor
growth as shown in fig.5 [3]Mohamed Lamine Toure“Advanced Algorithm for Brain
Segmentation using Fuzzy to Localize Cancer and Epilepsy
Region”, International Conference on Electronics and
Information Engineering (ICEIE 2010), Vol. no 2.
[4] Dr.G.Padmavathi, Mr.M.Muthukumar and Mr. Suresh
Kumar Thakur, “Non linear Image segmentation using fuzzy c
means clustering method with thresholding for underwater
images”, IJCSI International Journal of Computer Science
Issues, Vol. 7, Issue 3, No 9, May 2010
[5] Matei Mancas, Bernard Gosselin, Benoît macq,
“Segmentation Using a Region Growing Thresholding”
[6] S.Karpagam and S. Gowri “Detection of Glioma (Tumor)
Growth by Advanced Diameter Technique Using MRI Data”
Fig.5 tumor growth
CONCLUSION
We have presented a new method for Brain Segmentation to
localize Tumour. The method of segmentation of colour
images is based on fuzzy classification. It uses a fine initial
segmentation obtained by applying the FCM algorithm based
90
All Rights Reserved © 2012 IJARCSEE