IRJET- Brain Tumor Detection and Identification using Support Vector Machine

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 12 | Dec 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 2020
Brain Tumor Detection and Identification using Support
Vector Machine
Vinay J. Nagalkar1, Dr. G.G. Sarate2
1Assistant Professor, Dept of Electronics and Tele., VPKBIET Baramati, Maharashtra, India
2Professor, Dept. of Electronics and Tele. Government Polytechnic Amravati, Maharashtra, India
----------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - In this paper the work is related to the detection
and identification of brain tumor with the help of LBF SVM.
Proposed system helps doctors to find its region, size and
identify the disease. The brain software is having embedded
with graphical user interface for the tranquil and contented
use of the system by a non-technical person. Proposed system
is designed and developed for radiologist and neurologist.
After experimental result, we conclude that implemented
Brain Tumor identification strategy with LBF SVM optimize
the performance of doctors for detecting the brain disease
meritoriously. Proposed system experimental result tested
after rigorous experimentation, and the results are evaluated
on the basis of evaluation metrics,
Key Words: SVM (Support Vector Machine), LF SVM(Linear
Function).
1. INTRODUCTION
The brain disease brain tumor is required to
quantify as early as possible right from first stage of its
occurrences. The Neurologist takes help regarding the
quantification & rectification of such disease from the
radiologist. In this aspect radiologist is the expert for taking
the images of brain and also rectifies the data, which is
forwarded to neurologist for treatment. In this paper a
methodology to detect, quantify and also identify brain
disease using the digital image processing is presented. The
results are shown in order to clarify the methodology &
experimental outputs which will be used by the radiologist,
thus giving a second opinion to doctors for making the
decisions.
1.1 INDICATION OF BRAIN TUMOR
A brain tumour is nothing but abnormal growth of
cells in the brain. There are different types of brain benign
which are non-cancerous tumourtypeandsecondmalignant
tumour which is a cancerous type. The change to a high-
review tumour happens more regularly among grown-ups
than kids. The CT scan images of the tumour show the white
spots in the middle of the grey matter when the tumour is of
Grade I, and it is impossible to find this tumour in the brain
by using the radiologist expertise as shown in figure 1.
Fig -1: CT scanned images of brain tumor formed in
human brain.
2. LITERATURE REVIEW
The conventional strategy for distinguishing the
tumour ailments in the human MRI cerebrum pictures is
done physically by a doctor. Programmed grouping of
tumours of MRI pictures requires high precision, since the
non-exact finding and putting off conveyance of the exact
determination would prompt increment the commonness of
more certain sicknesses. To maintain a strategic distance
from that, a programmed arrangement framework is
proposed for tumour order of MRI pictures. Their work
demonstrates the impact of neural system (NN) and K-
Nearest Neighbor (K-NN) calculations on tumour
arrangement. The present study utilized a benchmark
dataset MRI mind pictures. The exploratory outcomes
demonstratethatusedinpresentmethodologyaccomplishes
100% arrangement exactness utilizing KNN and 98.92%
utilizing NN [1].
A novel calculation was done by Bhattacharjee and
Chakraborty (2012) to include out the tumorfromsick mind
with the help of Magnetic Resonance (MR) pictures. The
research enlightens the value parameter correlation of two
channels, a versatile middle channel is chosenforde-noising
the pictures. Picture cutting and distinguishing proof of
critical planes are finished. Coherent activities has been
connected on chosen cuts to acquire the handled picture
demonstrating the tumour locale. An epic picture remaking
calculation was created dependent on the use of Principal
Components Analysis (PCA) [2].
Chandra, Bhat and Singh (2009) has proposed a
grouping calculation dependent on Particle Swarm
Optimization (PSO). The calculation finds the centroids of
many bunches, where each group clusters together
cerebrum tumour designs, got from MR Images [3].
The research propsed by Dvorak, Kropatsch and
Bartusek (2013) has been manages programmed mind

tumour location in attractive resonant pictures. The
identification depends on checking the left-right symmetry
of the cerebrum, which is the suspicion for a solid mind. The
calculation was tried by fivefold cross-approval system on
72 pictures of mind containing tumours and 131 images of
the sound cerebrum. The proposed strategy achieves the
true positive rate of 91.16% and the true negative rate of
94.68% [4].
In present investigation, a digital method for
defining tumour region using Magnetic Resonance Imaging
(MRI) images in the brain was presented.A classificationofa
tumor is done depending upon the normal and abnormal
then further they have classified into a malignant and
benignant tumour. It incorporates different algorithms for
preprocessing, image segmentation, feature extraction and
image classification using neural network techniques.
Finally, the tumor area is specified by region of interest
technique as a confirmation step [5].
3. PROPOSED SYSTEM DESIGN
The dataset of CT Scan images of normal and
abnormal images are collected. Total of 1.5 lakh images are
collected from which 70:30 ratio is used for training the
linear function support vector machine. Thefollowingfigure
2 shows the proposed system design of the software
developed.
Fig - 2: Block diagram for training of classifiers.
The input database image is preprocessed. Preprocessing
like resizing, grayscale conversion, median filtering, image
enhancement is carried out on the images. The process is
shown in below block diagram figure 3.
Fig – 3: Block diagram of preprocessing.
Next stage is to extract the features from the database
images. In all for the quantification of brain tumor 14
features are extracted from which 4 features shows the
prominent change which can be used fortheidentification of
brain tumor. The training of the classifiers is done with the
database processed image and out of bag classification is
done as shown in figure 4.
Fig – 4: Testing of classifiers block diagram.
4. EXPERIMENTAL RESULT AND DESIGN
The classifiersusedinthe researchwork areLFSVM
which is trained by using the 30 normal images of human
brain and 70 different images having brain images. The
identification of the normal images given as an input to
classifiers 60 such images are usedtocheck theperformance
of the classifiers. The detection of all the 60 normal CT scan
images is detected as normal brain having no diseases in the
brain. The accuracy and precisionrateonthisinputdatabase
image is nearly 100%. First 30 images are used in the
training itself and another 30 images are out of bag
classification as shown in table 1

Table -1: LF SVM Results on normal brain images.
Images LF SVM Other Classifiers
Normal Image Normal Image
Normal Image Abnormal Image
The system is checked with the abnormal images having
brain tumor from the dataset. The results are tabulated as
shown below in table 2. The classification and identification
by LFSVM on such images are accurate and precise than any
other classifier. Confusion matrix is generated for the
calculation of various values for evaluation like Accuracy,
True Positive Rate, False Positive Rate, ErrorRate,Precision,
Recall, Sensitivity and F-Score of LF SVM as shown in figure
5
Fig - 5: Total system performance confusion matrix for LF
SVM classifier.
Table - 2: LF SVM Results on abnormal brain images.
Images LF
SVM
Detection of
Brain Tumor
Brain
tumor
Brain
tumor
Brain
tumor
Brain
tumor
Brain
tumor
5. CONCLUSION
Experimental result shows the accuracy rate of correctly
identifying the brain tumour in the input CT scan image
using Linear function Support VectorMachine32imagesout
of the 50 images are correctly classified giving an overall
accuracy of classifier as 64%. The time required for the
detection shows the detection and identification time of
classifier to find the disease. Time taken by LF SVM to detect
the tumour is 0.3525 seconds. The proposed system with
Linear Function Support Vector Machine classifier gives
better and perfect result to doctors for detecting and
discerning the brain disease.
REFERENCES
[1] Al-Badarneh, Amer, Hassan Najadat, and Ali M. Alraziqi.
”A Classifier to Detect Tumor Disease in MRI Brain
Images.” 2012 IEEE/ACM International Conference on
Advances in Social Networks Analysis and Mining.
Istanbul, Turkey: IEEE, 2012. 784-787.
[2] Bhattacharjee, Rupsa, and Dr. Monisha Chakraborty.
”Brain Tumor Detection From MR Images: Image
Processing, Slicing and PCA Based Reconstruction.”
Third International Conference on Emerging

Applications of Information Technology. Kolkata, India:
IEEE, 2012. 97-101.
[3] Chandra, Satish, RajeshBhat,andHarinderSingh.”APSO
Based method for Detection of Brain Tumors from
MRI.”World Congress on Nature & Biologically Inspired
Computing. Coimbatore, India: IEEE, 2009. 666-671.
[4] Dvorak, Pavel, Walter Kropatsch, and Karel Bartusek.
”Automatic Detection of Brain Tumors in MR Images.”
36th International Conference on Telecommunications
and Signal Processing. Rome, Italy:IEEE,2013.577-580.
[5] Ehab F. Badran, Esraa Galal Mahmoud, Nadder Hamdy.
”An Algorithm for Detecting Brain Tumors in MRI.”
(IEEE ) 2010.
"Brain Tumor Detection From MR Images: Image
Processing, Slicing and PCA Based Reconstruction."
IEEE, 2012. 97-101.
[7] Chandra, Satish, Rajesh Bhat, and Harinder Singh. "A
PSO Based method for Detection of Brain Tumors from
MRI." World Congress on Nature & BiologicallyInspired
Computing. Coimbatore, India: IEEE, 2009. 666
[8] Doker, Tarek A. El. "Unsupervised Iterative
SegmentationandRecognitionofAnatomic Structuresin
Medical Imagery Using Second-Order B-Spline
Descriptors and Geometric Quasi-invariants." Third
IEEE Symposium onBioInformaticsandBioEngineering.
Bethesda, MD, USA,: IEEE, 2003. 1-7.
[9] Al-Badarneh, Amer, Hassan Najadat, and Ali M. Alraziqi.
"A Classifier to Detect Tumor Disease in MRI Brain
Images." 2012 IEEE/ACM International Conference on
Advances in Social Networks Analysis and Mining.
Istanbul, Turkey: IEEE, 2012. 784-787.
[10] Amitava Halder, ChandanGiri,andAmiya Halder."Brain
tumor detection using segmentation based Object
labeling algorithm." International Conference on
Electronics, Communication and Instrumentation.
Kolkata, India: IEEE, 2014. 1-4.
"Brain Tumor Detection From MR Images: Image
Processing, Slicing and PCA Based Reconstruction."
IEEE, 2012. 97-101.
[12] Chandra, Satish, Rajesh Bhat, and Harinder Singh. "A
PSO Based method for DetectionofBrainTumorsfrom
MRI." World CongressonNature&BiologicallyInspired
Computing. Coimbatore, India: IEEE, 2009. 666-671.
BIOGRAPHIES
Vinay J. Nagalkar
He has completed his B.E(EXTC)in
the year 2010 and M.E (D.E) in the
year 2012. He is currently working
as Assistant professor at VPKBIET
Baramati and also he is Ph.D
Scholar from SGBAU Amarvati. He
has won Dr. S.K. Mukherjee
National Research Award for his
work in Digital Image Processing
and several state research awards.
Dr. Gajanan G. Sarate
He has completed his Ph.D. in
Optical Fiber Devices in 2006. He
has 28 years of teaching
experience. He is Ph.D. supervisor
in RTM Nagpur and SGB Amravati
University. He has guided 12 Ph.D
students out of which 8 has
successfully completedPh.D,2has
submitted Ph.D. thesis and 2
students are pursuing Ph.D under
him.

IRJET- Brain Tumor Detection and Identification using Support Vector Machine

More Related Content

What's hot (20)

Similar to IRJET- Brain Tumor Detection and Identification using Support Vector Machine (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Brain Tumor Detection and Identification using Support Vector Machine