BRAIN TUMOR DETECTION USING

SEGMENTATION OF MRI
Supriya Hinge, Anurag Thoke, Rishav Verma, Sohamniranjan Padvi, Dr. Meenakshi Thalor
hingesupriya@gmail.com, smpd11@gmail.com, meenakshithalor@gmail.com
Student, Computer Engineering, AISSMS’s IOIT, Savitribai Phule Pune University, Pune

ABSTRACT
Image segmentation is a key step from the image processing 3. REVIEW OF EXISTING
to image analysis. Based on preprocessing, segmentation and
extracting the features it converts the original image to a
APPROACHES
compact form, and with the help of this, it is possible to make Paper Title Year Approach Limitations
proper image analysis and understanding the image. If the Brain Tumor 2016 Novel CNN- They should
MRI is given as an input, it can be converted into a gray scale Segmentation.U based method first be
image and with the help of various other algorithms, Brain sing for trained using
tumor can be detected. Convolution segmentation Learning
Neural of brain process,
General Terms Networks in tumors in MRI which
Segmentation, Classification, PreProcessing MRI Images images sometimes
takes very
Keywords long
Denoising, Skull Striping, Intensity Normalization, Benign, Efficient 2016 MRI is The
Malignant Detection of segmented algorithm
. Brain Tumor using k- could not find
from MRIs clustering out the
1. INTRODUCTION Using K-Means algorithm, precise or
Bio Medical Processing of image is a growing field. There are Segmentation SVM and accurate
various types of medical imaging processes available which and Normalized Naïve Bayes boundary of
comprise of many different types of imaging like CT scans, Histogram Approach the tumor
X-Ray and MRI. Magnetic Resonance Image(MRI) is the region.
most reliable and safe. MRI can be preprocessed, and the A Survey on 2017 Fuzzy C Sample
image can be segmented. This whole process can be done Brain Tumor Means selection and
with the help of Image Processing. The process includes: Detection Using establishing
PreProcessing, Segmentation, Optimization and Feature Image fuzzy sets
Extraction followed by classification, size and volume Processing may be
detection and stage detection. These techniques allow us to Techniques tedious
identify even the smallest abnormalities in the human brain. A Two Phase 2016 Phase I- Only T1
The goal of medical imaging is to extract accurate information Segmentation Histogram images are
from these images with the least error possible. Manual Algorithm For Thresholding considered.
Segmentation of human brain tumor is a very tedious job and MRI Brain Phase II- Does not
also time consuming. Hence, Image Processing techniques are Tumor Region work on other
applied nowadays in Medical Imaging. Extraction Growing type.
phase
2. BACKGROUND
Review Of 2013 Combination Noise may
Magnetic resonance imaging (MRI) of the brain is a safe test
Brain Tumor of both lead to
that is carried out using a magnetic field and radio waves to
Detection Using modified undesired
produce detailed images of the brain. MRI is used in detecting
MRI Images texture based artifacts in
various conditions of the brain such as swelling, cysts, tumors,
region final result
bleeding, infections, and inflammatory conditions. The MRI is
growing and
then analyzed by a radiologist who is trained in interpreting
cellular
the scans. The Radiologist sends a report to your doctor, who
automata edge
then discusses the results with you and explain what kind or
detection is
problem is. The scan includes 9 slices of MRI scans, it takes a
used
lot time to evaluate what type of tumor is and the other
Detection of 2016 Median filter High
characteristics of tumor, so to overcome that problem, direct
Brain Tumor and by using computationa
evaluation of MRI preprocessed images is done in our project.
from MRI diagonal, l cost

Volume 3 Issue 2 April 2018 163

images by using antidiagonal existing tool known as 3D Slicer which converts a
Segmentation masks 2D image into a 3D image. The second approach is
&SVM segmented Connected Components method.
images. 7. Stage Detection: Finally, the Stage of the tumor is
Malignant 2012 GVF works on The detected using the size, volume and type of tumor as
Brain Tumor the principle parameters of parameters.
Detection of energy GVF has to
minimization be controlled
hence manually
effectively which is time
extracting the consuming
tumor. process and
may lead to
error in the
results
Brain Tumor 2016 K-Nearest Possibility of
Detection Using Neighbor yielding an
Pattern (KNN) erroneous
Recognition decision if
Techniques obtained
single
neighbor is
an outlier of
some other
class
Brain Tumor 2015 Meyer's Cannot be
Extraction from flooding used for
MRI Images Watershed images with
Using Algorithm poor contrast Figure 1. Architecture of Proposed System
MATLAB or images
with a lot of
background
5. ALGORITHMS AND PSEUDO CODE
and
foreground PreProcessing: Sobel-Feldman Operator
artifacts Mat srcImage;
Mat mRgbMat = new Mat();
Mat mHsvMat = new Mat();
Mat mMaskMat = new Mat();
4. PROPOSED SYSTEM Mat mDilatedMat = new Mat();
The proposed system consists of seven modules: Pre- Mat hierarchy = new Mat();
Processing, Segmentation, Feature Extraction, Classification,
Size Detection, Volume Detection, Stage Detection. Segmentation: Thresholding
1. Pre-Processing: Pre-Processing is carried out by Scalar lowerThreshold = new Scalar(0, 0, 99);
filtering using Sobel-Feldman Operator. It is also Scalar upperThreshold = new Scalar(255, 10, 255);
known as the Sobel Operator or Sobel filter. This
filter is used in processing of image, especially for Feature Extraction: Histogram Oriented Gradient
edge detection techniques, where the focus is on Imgproc.cvtColor(mRgbMat,mHsvMat,
emphasizing the edges. Imgproc.COLOR_RGB2HSV_FULL );
2. Segmentation: The process of segmentation is done
by using Thresholding. It is the simplest form of Size Detection: Bounding Box Method
image segmentation. The final output of this stage is for (int i = 0; i < tumorHeigh.size(); i++)
binary images. {
3. Feature Extraction: Feature Extraction is done by dlm.addElement(tumorWidt.get(i) + " X " +
Histogram Oriented Gradients (HOG). This is tumorHeigh.get(i));
basically like a feature descriptor used for the }
purpose of object detection. if (area > 200 && area < 1200)
4. Classification: In the next step, Classification is {
done using K-Nearest Neighbor. The k-NN Imgproc.drawContours(mRgbMat, contours, -1,
algorithm is a non parametric method used for the new Scalar(40, 233, 45, 0), 4);
purpose of classification and also regression. Imgproc.rectangle(inputImage, rct.tl(), rct.br(),
5. Size Detection: After this classification, the size new Scalar(255, 0, 0, 255), 2);
and volume of the Tumor is calculated using }
Bounding Box Method.
6. Volume Detection: This can be carried out using
two different approaches, the first one being an

Volume 3 Issue 2 April 2018 164

Classification: K-Nearest Neighbor
for (int i = 0; i < tumorHeigh.size(); i++)
{
if (Integer.parseInt("" + tumorWidt.get(i)) >= 10 &&
Integer.parseInt("" + tumorWidt.get(i)) <= scale1)
{
stage = "Benign";
} Figure 4: Output of Thresholding
else if (Integer.parseInt("" + tumorWidt.get(i)) >= scale2 &&
Integer.parseInt("" + tumorWidt.get(i)) <= scale3)
{
stage = "Malignant";
}
}
Figure 5: Output of Histogram Oriented Gradient
6. RESULTS
Tools:
OpenCV
OpenCV (Open Source Computer Vision Library) is an open
source computer vision and machine learning software library.

NetBeans NetBeans Figure 6: Output of Classification

IDE lets you quickly and easily develop Java desktop, mobile
and web applications.

Java
Java is used in a wide variety of computing
platforms from embeddeddevices and mobile
phones to enterprise servers and supercomputers.
Figure 7: Output of Size Detection
SQL
SQL is a language to operate databases; it includes database
creation, deletion, fetching rows, modifying rows, etc.

Fig.2. to Fig. 8 are the snapshots of real time system.

Figure 8: Final Result

7. CONCLUSION AND FUTURE WORK
To examine the location of tumor in the brain, MRI is used.
Radiologists will evaluate the grey scale MRI images, but this
procedure is really time and energy consuming.
As a general conclusion, it can be summarized that the main
Figure 2: GUI objective of this project is to develop a technique which not
only reduces the efforts of a radiologist, but also assists in
detection of tumor.
A novel algorithm for the detection of tumor in brain is
described in this research. Our approach successfully
managed to depict that the system proposed is a valuable
diagnosis technique for the radiologists to detect the brain
tumors.
In future, additional information can be included about the
Figure 3: Output of Sobel Feldman Operator features. It will be interesting to continue developing more
adaptive methods for other types of brain tumors following
the same approach. Another future task would be the detection
of other factors which influence the appearance of tumors on
images and though there are some features which are common
of malignant and benign tumors, there is a great amount of
variation that depends on the tissue and tumor type. Efforts

Volume 3 Issue 2 April 2018 165

can be made to reduce some effects such as architectural [7] Amarjot Singh, Shivesh Bajpai, Srikrishna Karanam,
distortion. Akash Choubey and Thaluru Raviteja “Malignant Brain
Tumor Detection International Journal Of Computer Theory
And Engineering”, International Journal of Computer Theory
8. REFERENCES and Engineering, Vol. 4, No. 6, December 2012.
[1] Sergio Pereira, Adriano Pinto, Victor Alves, And Carlos
[8] Bandana Sharma, Dr. Brij Mohan Singh Bandana Sharma
A. Silva “Brain Tumor Segmentation Using Convolutional
Et Al, “Brain Tumor Detection Using Pattern Recognition
Neural Networks In MRI Images” IEEE Transactions on
Techniques”, International Journal Of Recent Research
Medical Imaging Volume: 35, Issue: 5, May 2016 .
Aspects ISSN: 2349-7688, Special Issue: Conscientious And
[2] Garima Singh Dr. M.A. Ansari, “Efficient Detection Of
Unimpeachable Technologies 2016 .
Brain Tumor From MRI Using K-Means Segmentation And
[9] Rajesh C. Patil, Dr. A. S. Bhalchandra “Brain Tumor
Normalized Histogram”,1st India International Conference on
Extraction From MRI Images Using MATLAB” International
Information Processing (IICIP) ,2016
Journal Of Electronics, Communication Soft Computing
[3] Luxit Kapoor, “A Survey On Brain Tumor Detection
Science And Engineering ISSN: 2277-9477, Volume 2, Issue 1
Using Image Processing Techniques”, Amity School Of
,April 2012.
Engineering And Technology Amity University, Noida ,India,
[10]M.Raghavi, M.Princy, R.Priyanka, Mrs.A.Lakshmi, “3D
2017 IEEE.
Volume calculation of Brain Tumor Using HOG Feature
[4] R. Anita Jasmine, Dr.P. Arockia Jansi Rani, “A Two
Extraction And Connected Component”, International
Phase Segmentation Algorithm For MRI Brain Tumor
Journal Of Digital Communication And Networks(IJDCN)
Extraction”, International Conference on Control,
Instrumentation, Communication and Computational Volume 1,Issue 3,September 2014.
Technologies, 2016 .
[5] Miss Hemangi S. Phalak, Mr. O. K. Firke Review Of
“Brain Tumor Detection Using MRI Images” International
Journal For Research In Applied Science Engineering
Technology (IJRASET) Volume 4 Issue III, March 2016 .
[6]Swapnil R. Telrandhe Amit Pimpalkar Ankita Kendhe,
“Detection Of Brain Tumor From MRI Images By Using
Segmentation SVM”, World Conference On Futuristic Trends
In Research And Innovation For Social Welfare (Wcftr16),
2016.
.

Volume 3 Issue 2 April 2018 166