See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/310464150

DETECTION OF BRAIN TUMOR FROM MRI OF BRAIN

Article · December 2015

CITATIONS READS

3 3,913

3 authors, including:

Madhurima Banerjee Samir Bandyopadhyay

The Heritage Academy University of Calcutta
12 PUBLICATIONS 7 CITATIONS 1,076 PUBLICATIONS 11,990 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Madhurima Banerjee on 18 November 2016.

The user has requested enhancement of the downloaded file.

 International Journal of Information Research and Review, December 2015

International Journal of Information Research and Review

Vol. 02, Issue, 12, pp. 1555-1559 December, 2015

Research Article
DETECTION OF BRAIN TUMOR FROM MRI OF BRAIN
1Madhurima Banerjee, 2Ranjita Chowdhury and 3*Samir Kumar Bandyopadhyay
1Department of Computer Science and Application, Heritage Academy, Kolkata, India
2Department of computer Science and Engineering, St. Thomas college of Engineering and Technology, Kolkata,
India
3Department of Computer Science and Engineering, University of Calcutta, Kolkata, India

ARTICLE INFO ABSTRACT

Article History: Tumor segmentation from magnetic resonance imaging (MRI) data is an important but time
consuming manual task performed by medical experts. Magnetic resonance imaging (MRI), computed
Received 09th, September 2015 tomography (CT), digital mammography, and other imaging processes give an efficient means for
Received in revised form
detecting different type of diseases. Here a method is proposed for detection of brain tumor from MRI
19th, October 2015
Accepted 26h, November 2015 of brain.
Published online 30th, December 2015

Keywords:
MRI of Brain,
Automated System,
Tumor Segmentation,
Tumor Detection,
Automated System,
Pre-processing,
Filtering,
Features Extraction.

Copyright © 2015 Madhurima Banerjee et al. This 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.

INTRODUCTION
different type of MRI images, finally, the abnormal portion is
The automated detection methodology have deeply improved extracted (image segmentation) to perform volumetric analysis
knowledge of normal and diseased examination for medical and classified for treatment planning varies for different types
research and are a important part in diagnosis and treatment of abnormalities which verify the success rate of the treatment
planning when the number of patients increases (TamijeSelvy given to the patient. Standard x-rays and computed tomography
et al., 2011). Among all the medical image modalities, (CT) can initially be used in the diagnostic process. However,
Magnetic Resonance Imaging (MRI) is the most frequently MRI is generally more useful because it provides more detailed
used imaging technique in neuroscience and neurosurgery for information about tumor type, position and size. For this
these applications. The brain abnormality detection and reason, MRI is the imaging study of choice for the diagnostic
segmentation on MRI images is a very difficult and vital task work up and, thereafter, for surgery and monitoring treatment
which is used in surgical and medical planning and assessment. outcomes (Laurence et al., 1998).
The difficulty in brain image analysis is mainly due to the
requirement of detection techniques with high accuracy within Brief Review
quick convergence time. Automating this process is a
challenging task because of the high diversity in the appearance A magnetic resonance imaging instrument or MRI Scanner
of abnormal tissues among different patients and in many cases (Magnetic Resonance Imaging, 1998) uses powerful magnets to
almost similar with the normal tissues. The detection process of polarize and excite hydrogen nuclei i.e. proton in water
any abnormalities in the brain images are a two-step process. molecules in human tissue, producing a detectable signal which
Initially, the abnormal MR brain images are detected from is spatially encoded, resulting in images of the body (Aria
Tzika et al., 2011). MRI mainly uses three electromagnetic
fields they are : i) A very strong static magnetic field to
*Corresponding author: Samir Kumar Bandyopadhyay
polarize the hydrogen nuclei, named as the static field, ii) A
Department of Computer Science and Engineering, University of
Calcutta, Kolkata, India. weaker time varying field(s) for spatial encoding, named as the
1556 Madhurima Banerjee et al. Detection of brain tumor from mri of brain

gradient field, iii) A weak radio frequency field for or malignant tumor and other different considerations. Brains
manipulation of hydrogen nuclei to produce measurable signals tumors are the solid portion permeate the surrounding tissues or
collected through RF antenna. The variable behaviour of distort the surrounding structures (Roger et al., 2002). There are
protons within different tissues leads to differences in tissue different type of brain tumor they are i) Gliomas, ii)
appearance. The different positioning of MRI of brain with T1 Medulloblastoma, iii) Lymphoma, iv) Meningioma, v)
and T2 weight is shown below. (Figure 1, igure 2 and Figure 3) Craniopharyngioma, vi) Pituitary adenoma.

(a) (b) (c) (d) (e) (f)

Figure 1. MRI of brain cited by http://www.mr-tip.com/serv1.php?type=isimg. T2 weighted MR image (a) brain shows cortex,
lateral ventricle, and falxcerebri, (b) brain shows eyeballs with optic nerve, medulla, vermis, and temporal lobes with hippocampal
regions, (c) head shows maxillary sinus, nasal septum, clivus, inner ear, medulla, and cerebellum. T1 weighted MR image (d) brain
shows cortex, white and grey matter, third and lateral venticles, putamen, frontal sinus and superior sagittal sinus, (e) brain shows
eyeballs with optic nerve, medulla, vermis, and temporal lobes with hippocampal regions,(f) brain shows cortex with white and grey
matter, corpus callosum, lateral ventricle, thalamus, pons and cerebellum from the same patients

(a) (b) (c) (d) (e)

Figure 2. a) Axial T1-weighted with tumor, b) T2-weighted with central positioning tumor, c) Contrast enhanced T1-weighted image
showing ring formed tumor, d) Contrast enhanced T1-weighted image with high grade oligodendro glioma e) T2-weighted image
with high grade oligodendro glioma from the same patient

(a) (b) (c) (d)

(e) (f) (g) (h)

Figure 2. a) original MRI image, b) original MRI with 50% noise, c) median filter, d) Min-Max Median Filter, e) Center
Weighted Median Filter, f) Adaptive Median Filter, g) Progressive Switching Median Filter, h) Average Filter

A brain tumor is a mass of cells that have grown and multiplied Proposed Method
uncontrollable i.e. a brain tumor is an uncontrolled growth of
solid mass formed by undesired cells either normally found in Automated system (detection) of brain tumor through MRI is
the different part of the brain such as glial cells, neurons, basically called Computer-Aided Diagnosis (CAD) system. The
lymphatic tissue, blood vessels, pituitary and pineal gland, CAD system can provide highly accurate reconstruction of the
skull, or spread from cancers mainly located in other organs original image i.e. the valuable outlook and accuracy of earlier
(The Essential Guide to Brain Tumors, 2012). Brain tumors are brain tumor detection. It consists of two or more stage. In the
classified based on the type of tissue involved in the brain, the initial stage pre-processing has required after that stages post-
positioning of the tumor in the brain, whether it is benign tumor processing i.e. segmentation are required.
1557 International Journal of Information Research and Review Vol. 02, Issue, 12, pp. 1555-1559, December, 2015

Then detection strategies and other information, feature image may be inadequate to maintain fine description then
extraction, feature selection, classification, and performance pseudo colour conversion are need. Each gray value maps to an
analysis are compared and studied. Pre-processing techniques RGB item. To obtain more useful features and enhance the
are used to improvement of image quality and remove small visual density, some of the method may applies pseudo-colour
artefacts and noise for the accurate detection of the undesired transformation, a mapping function that maps a gray-level pixel
regions in MRI. Post-processing is used to segment with to a colour level pixel by a lookup table in a predefined colour
different strategy the brain tumor from the MRI of brain map. An RGB colour map contains R, G, and B values for each
images. In this paper, here focus on the appearance of tumors in item. But most of the segmentation applied on the gray image
MRI images, the grade of tumors and some general information the RGB to gray conversion is needed.
which will be useful in the detection, segmentation and
interpretation of brain tumors from MRI images. The basic Algorithm
system is made up of individual modules and each individual
module is associated with its own techniques. The major Input: MRI of Brain Image.
modules of this work are image database, pre-processing,
feature extraction, classification and image segmentation. The Output: Binarizes MRI of Brain Image.
MRI slides can be gathered from several diagnostic centers and
hospitals. For experimental analysis images available in the Step1: Take an MRI image I(x,y).
public domain are utilized that are utilized by several research Step2: If it is color image then convert it into gray scale image
organizations those are conducting similar research. Pre- Ig(x,y).
processing mainly involves those operations that are normally Step3: Calculate standard deviation of the image and store the
necessarily prior to the main goal analysis and extraction of the intensity value in TS.
desired information and normally geometric corrections of the Step4: calculate the threshold value by product of standard
original actual image. These improvements include correcting deviation and a predefine constant H, i.e.
the data for irregularities and unwanted atmospheric noise, Threshold intensity value T= TS*H.
removal of non-brain element image and converting the data so Step5: Scan left to right and top to bottom, each pixel of the
they correctly reflected in the original image. gray image Ig(x,y).
Step6: Find a binary image IB from the gray image lg(x,y) in the
Segmentation is the process of partitioning an image to several following way,
segments but the main difficulties in segmenting an images are
i) Noise, ii) Blur Low Contrast, iii) The bias field (the IB(x,y) = 1 lg(x,y) >= T
occurrence of smoothly varying intensities within tissues) , iv) IB(x,y) = 0 lg(x,y) < T
The partial-volume effect (a voxel contributes in multiple tissue
types) . Image filtering and enhancement stage is the most Step7: IB is the output binary image.
obvious part of medical image processing. This pre-processing
stage is used for reducing image noise, highlighting important We use binarization technique of MRI of brain that so MRI of
portions, or displaying obvious portions of digital images. brain is used as an input. As the binarization technique can be
Some more techniques can employ medical image processing applied only to grayscale images. We convert RGB image to its
of coherent echo signals prior to image generation and some of corresponding grayscale image. A RGB image has three
the images are hanging from clip hence they may produce components red, green and blue and converts it into on
noise. The enhancement stage includes resolution enhancement; component i.e. gray value which lies between 0 to 255 intensity
contrast enhancement. These are used to suppress noise and values.
imaging of spectral parameters. After this stage the medical
image is converted into standard image without noise, film Then we calculate the standard deviation of the matrix elements
artefacts and labels. Methods such as use of standard filters to (image pixels).Thus by using standard deviation we select the
more advanced filters, nonlinear filtering methods, anisotropic random intensity values as the standard deviation values will be
nonlinear diffusion filtering, a Markov random field (MRF) less than 100 and hence we multiplied the deviated value by a
models, wavelet models, non-local means models (NL-means) constant value. Here we choose this constant value H=3.
and analytically correction schemes. These methods are almost Although H=3 is choosen, in few images H= 2.5 also produce
same in terms of computation cost, de-noising, quality of de- good results .Here we also gives a comparative study why we
noising and boundary preserving. choose constant H equal to 3.

Here we use visual inspection as well as quantative

In the proposed methods we used standard deviation to select
measurement to choose the constant. Visual inspection may be
the threshold intensity of the image. Ultimate selection of
biased but together with quantative measurement [8] such as
threshold has done by multiplying a constant value with the
ME, RAE, Precision, Recall, F-measure and visual are very
threshold intensity of the image using standard
effective. Thus after getting the threshold intensity we compare
deviation.Colour models or colour spaces, indicate the colours
each pixel of the gray image to find out whether it is greater
in a benchmark way by using a coordinate system and a
than or less than the threshold intensity value.
subspace in which each colour is represented by a single point
of the coordinate system. The largely common colour spaces
If the pixel intensity is greater than the threshold value then that
used in image processing [24, 25] methods are Gray, Binary
pixel value is set to 1 otherwise it is set to 0. Thus the whole
form, RGB, HSV, HIS etc. Gray form are used in all binarized
image is transformed into 0 or 1 i.e. a binary image is generated
method, i.e. most of the segmentation technique used
from the gray image where the foregrounds are marked as 1 and
binarization methods original MR Brain image is a gray-level
backgrounds are marked as 0.
1558 Madhurima Banerjee et al. Detection of brain tumor from mri of brain

Algorithm for abnormal region identification Step11: Label the connected components of image I with 8
connected neighbourhoods. Let it be Ilabel. Step12: Measure
Input: A MRI image of Brain. properties of image regions of labelled image Ilabel and compute
Output: Abnormal region identification with centroid number of pixels inregions.
Step1: Read Image (Gray image or Colour Image) Step13: Remove the components from Labelled image Ilabel is
Step2: Convert it to Gray Image (IGRAY) form RGB. not a member and having pixels less than some predefined
Step3: Remove noise by using median filter of image IGRAY. value. Let the processed image is H.
The processing Image is Ipro.
Step4: Set L=Ipro Step14: Create „Sobel‟ horizontal edge emphasizing filter (2-D
Step5: Compute standard deviation of Ipro using formula above. filter) predefined as a correlation kernel.
Let it .
Step6: Using this perform intensity map as follows. Step15: Compute the gradient magnitude of image H as
[m , n]= size[L] // size of image is calculated
follows. Gradient magnitude=
For i=1 to m
Where Ix is obtained by filtering the image H by „Sobel‟ filter
For j=1 to n created in Step14. Iy is obtained by filtering image H by
If L(i,j) > Then transpose of „Sobel‟ filter created in step14.

setL(i,j)=255 Step16: The step 13 gives the Tumor region and Step15 gives
the border of Tumor region.
Else
setL(i,j)=0 RESULTS
End If The proposed algorithm is used to detect the location or
position the abnormal regions, centroid of the abnormal
End For regions, and border of the abnormal regions and the perimeter
of the detected abnormal regions. The output of the segmented
End For
abnormal regions and border of the Abnormal regions are plays
Step7: Recompute the standard deviation of image L. Let it
a vital role in the diagnosis and treatment planning of brain
. abnormal regions. The proposed method gives very reasonable
Step8: Compute average intensity of the pixels those are above results for different kind of MRI images.
the as follows.
Set count 0 Conclusion
Set Isum 0
For i=1 to m The proposed algorithm shows an effective method for
segmentation of the brain tumors from the 2 dimensional MRI
For j=1 to n images. We have proposed to detect presence of brain tumor
If I(i,j) Then based on thresholding technique. The segmentation of the brain
is also being done while detecting the presence of the tumor.
set count =count +1// pixel count. We also find the centroid of the tumor and perimeter of the
End If tumor. In this paper we have developed an automatic image
End For based method to detect tumors in 2-D MRI head scans.
Experimental results on 30 data sets show that the proposed
End For method performed Automatic Detection of Brain Tumor from
MRI Scans. Future research scope in segmentation of medical
Step9: Average intensity, Iaverage =Isum/count. // average intensity
MRI image will lead towards improving the accuracy,
calculation. Step10: Take this average intensity as Threshold
minimizing the computational procedure also minimize the
value to find tumor section.
manual interaction.
For i=1 to m
REFERENCES
For j=1 to n “Magnetic Resonance Imaging (MRI) Scans and Procedures”,
Diagnostic Imaging Pathways. http://www.imagingpath
If I(i,j) >Iaverage Then set I(i,j)=255 ways.health. wa.gov.au/includes/pdf/consumer/mri.pdf
“The Essential Guide to Brain Tumors” National Brain tumor
Else society. http://www.braintumor.org/patients-family-friends/
setI(i,j)=0; about-brain-tumors/publications/essentialguide.pdf
Aria Tzika, LoukasAstrakas and Maria Zarifi ,“Pediatric Brain
End If Tumors: Magnetic Resonance Spectroscopic Imaging”
End For Diagnostic Techniques and Surgical Management of Brain
Tumors, Department of Surgery, Massachusetts General
End For Hospital, Harvard Medical School, Boston, USA, pp- (205-
226).
1559 International Journal of Information Research and Review Vol. 02, Issue, 12, pp. 1555-1559, December, 2015

Laurence, P. Clarke, Robert, P. ?. Velthuizen, Matt Clark, Roger J. Packer, Henry S. Friedman, Larry E. Kun, And
Jorge Gaviria, Larry Hall, Dmitry Goldgof, ReedMurtagh, Gregory N. Fuller “Tumors Of The Brain Stem,
S. Phuphanich, Steven Brem 1998. “MRI Measurement of Cerebellum, Fourth Ventricl”, 2/15/02 Page 171-192.
Brain Tumor Response: Comparison of Visual Metric and http://soc-neuro-onc.org/levin/Levin_ch06_p171-192.pdf
Automatic Segmentation”, Magnetic Resonance Imaging, TamijeSelvy, P., V. Palanisamy, T. 2011. Purusothaman,
Volume 16, Issue 3, Pages 271-279, April. “Performance Analysis of Clustering Algorithms in
Murtagh, S. Phuphanich, Steven Brem 1998. “MRI BrainTumor Detection of MR Images”, European Journal
Measurement of Brain Tumor Response: Comparison of of Scientific Research, Vol.62 No.3 pp. 321-330.
Visual Metric and Automatic Segmentation”, Magnetic
Resonance Imaging, Volume 16, Issue 3, Pages 271-279,
April.

*******

View publication stats