IRJET- Brain Tumor Detection using Convolutional Neural Network

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 01 | Jan 2020 www.irjet.net p-ISSN: 2395-0072
© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 1232
Brain Tumor Detection using Convolutional Neural Network
Sachin R Jadhav1,Shubham S Salve2, Harshal S Mohagaonkar 3, Akhilesh D Rakibe4,
Nishant G Langade5
1,2,3,4,5Department of Information Technology, Pimpri Chinchwad College of Engineering, Nigdi, Pune, Maharashtra
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract -Image processing is the one of the most demanding
and promising field nowadays. Tumoris aabnormalgrowthof
cell in human brain. The tumor can be categorized as
benign(non-cancerous) and malignant(cancerous). Earlier
stage of tumor is used to be detected manually through
observation of image by doctors and it takes more time and
sometimes gets inaccurate results. Today different automated
tools are used in medical field. These tools provide a quick and
precise result. Magnetic Resonance Images (MRI) is the most
widely used imaging technique for analyzing internal
structure of human body. The MRI is used even in diagnosis of
most severe disease of medical science like brain tumors. The
brain tumor detection process consist of image processing
techniques involves four stages. Image pre-processing, image
segmentation, feature extraction, and finally classification.
There are several existing of techniquesareavailableforbrain
tumor segmentation and classification to detect the brain
tumor. There are many techniques available presents a study
of existing techniques for brain tumor detection and their
advantages and limitations. To overcome these drawbacks,
propose a Convolution Neural Network(CNN)basedclassifier.
CNN based classifier used to compare the trained and test
data, from this get the best result.
Key Words: Brain Tumor Detection, CNN, Image Pre-
processing.
1. INTRODUCTION
The Image processing is a processofanalyzing,manipulating
an image in order to perform some operation to extract the
information from it. Medical imaging seeks to disclose
internal structures hidden by skin and bones and also to
diagnose and treat disease. And also, it establishes a
database of normal anatomy and physiology to make it
possible to identify abnormalities. In today’s world, one of
the reasons in the rise of mortality amongthepeopleisbrain
tumor. Abnormal or uncontrolled growth of cell developed
inside the human body is called brain tumor. This group of
tumor grows within the skull, due to which normal brain
activity is disturbed. Brain tumor is a serious life frightening
disease. So, which not detected in earlier stage, can take
away person’s life. Brain tumors can be mainly three
varieties called benign, malignant, pre-malignant. The
malignant tumor leads to cancer.
Treatment of brain tumor depends on many factors such as
proper diagnosis and the different factor like the type of
tumor, location, size, and state of development. Previously
stage of tumor is used to be detected manually with the help
of observation of image by doctors and sometimes it takes
more time and results may be inaccurate. There are many
types of brain tumor and only expert doctor can able to give
the accurate result. Today many computers added tool is
used in a medical field. These tools have a property of quick
and accurate result.
MRI is the most commonly used imaging technique for
inspecting internal structure of human body. Proper
detection of tumor is the solution for the proper treatment.
Also require accurate diagnosis tool for proper treatment.
Detection involves finding the presence of tumor. Detecting
brain tumor usingimageprocessingtechniquesinvolvesfour
stages. Image pre-processing, segmentation, feature
extraction, and classification. The primary task of pre-
processing is to improve the quality of the Magnetic
Resonance (MR) images, removing the irrelevant noise and
undesired parts in the background and preserving its edges.
In segmentation the pre-processed brain MR images is
converted into binary images. Feature extraction is the
process of collecting higher level information of an image
such as color, shape, texture and contrast. And the
classification process, the classifier is used to classify the
normal trained image samples and the input image sample.
1.1 LITERATURE SURVEY
[1] Capsule Networks for Brain Tumor Classification Based
On MRI Images And Coarse Tumor Boundaries.
As stated by the WHO, cancer is deemedto besecondleading
cause of human casualties. Out of different types of cancer,
brain tumor is perceived as one of the fatal due to its
vigorous nature, diverse characteristics and relatively low
survival rate. Discovering the type of brain tumor has
remarkable impact on the choice of therapy and patient’s
survival. Human based identification is usually inaccurate
and unreliable leading in a recent sweep of interest to
automize this process using convolutional neural network
(CNN). As CNN fails to completely utilize spatial relations,
which may lead to incorrect tumor classification. In our
technique, we have included newly evolved CapsNet to
prevail this shortcoming. The main offering is to provide
CapsNet with access to tissues neighbouring the tumor,
without diverting it from the principal target. An improved
CapsNet architecture is consequently proposed for the
classification of brain tumor, that takes the coarse
boundaries of tumor as additional input within its pipeline
for surging the focus of the CapsNet.
[2] A Hybrid Feature Extraction Method with Regularized
Extreme Learning Machine for Brain Tumor Classification

Classification of the brain tumor is the crucial step that
depends upon understanding and expertiseofthephysician.
The automated classification system of the brain tumor is
vital to assist radiologists and physicians to identify the
tumor. Nonetheless, the precision of the current systems
needs to be improved for the successful treatment. In this
paper the proposed approach consists of, (1) brain image
pre-processing, (2) feature extraction of the image & (3)
brain tumor classification. Initially the input images of the
brain are transformed into intensitybrainimagesusingmin-
max normalization rule resulting into enhanced and
improved contrast of the edges and regions of the brain.
Then by applying feature extraction to the brain images
using hybrid feature extraction and then computing the
covariance matrix of the features extracted to project them
into a notable set of features using principle component
analysis (PCA). Ultimately, the type of brain tumor is
classified using regularized extreme learning machine
(RELM). As per the results the suggestedapproachprovedto
be more effectual compared to the current approaches. Also
the performance in terms of accuracy of the classification
improved from 91.51% to 94.233% for the experiment.
[3] Tumor Detection and Classification of MRI Brain Image
using Different Wavelet Transforms and Support Vector
Machines
The brain is the principal organofhumanbody.Anabnormal
growth of cells leads to the brain tumor. This abnormal
growth of cells results in unusual functioning of brain and
eradication of healthy cells. The brain tumors can be
classified as malignant(cancerous) and benign(non-
cancerous) tumors. In this paper the proposed approach
includes (1) Pre-processing, (2) Training the SVM & (3)
Submit training set to SVM and output the obtained
predictions. At first stagedenoisingthemedical imagesusing
different kind of wavelets while maintaining the important
features. In segmentation for the extraction of the features,
Otsu method is used for converting grey-level image to
binary image. Finally, the data has two classes and we can
apply SVM for classification. The outcome shows that SVM
with proper training dataset is able to differentiate between
normal and abnormal tumor regions and categories as
malignant tumor, benign tumor or a healthy brain.
[4] Segmentation and Recovery of Pathological Mr Brain
Images Using TransformedLow-Rank andStructured Sparse
Decomposition
A general framework is proposed for the concurrent
segmentation and recovery of pathological magnetic
resonance images (MRI), where low rank and sparse
decomposition (LSD) schemes have been used extensively.
Due to the lack of constraint between low-rank and sparse
components, conventional LSD techniques often construct
recovered images with distorted pathological areas. For
resolving this issue, a transformed low rank and structured
sparse decomposition (TLS2D) method is proposed, that is
vigorous for taking out pathological regions. By using
structured sparse and computed image saliency as adaptive
sparsity constraint the well recovered images can be
acquired. The exploratory results ontheMRIimagesofbrain
tumor shows that the TLS2D can successfully provide
adequate performance on image recovery as well as tumor
segmentation.
[5] Brain Tumor Segmentation Using Convolutional Neural
Networks in MRI Images
Out of different types of brain tumors, malignant tumors are
assertive and commonly occurring, decreasing the life
expectancy. MRI is extensively used imaging method for
assessing the tumors. Due to the huge amount of data
produced by MRI stops the manual segmentation in a fair
time, restricting the use of accurate quantitative
measurements in clinical practices. For resolving this, an
automatic segmentation technique based on CNN is
proposed, exploring small kernels. Employing small kernels
allows designing a deeper architecture, alongside having an
advantage against overfitting, given the small number of
weights in the network. Use of intensity normalization in
pre-processing with data augmentation has proven to be
effectual for brain tumor segmentation in MRI images.
[6] Development of Automated Brain Tumor Identification
Using MRI Images
Brain tumor is a prime reason for human casualties every
year. Magnetic resonance imaging (MRI)isa commonlyused
techniquefor braintumordiagnosis.Anautomatedapproach
which incorporates enhancement at an early stagetoreduce
gray scale colour variations. For better segmentation the
unnecessary noises were decreased as much as possible
using filter operation. The proposed approach uses
threshold-based Otsu segmentation rather than colour
segmentation. Ultimately, the feature information provided
by the pathology experts was used to identify region of
interests. The exploratory results demonstrate that the
proposed approach was able to provide adequate results as
compared to present available approaches in terms of
accuracy.
[7] Brain Tumor Segmentation to Calculate Percentage
Tumor Using MRI
Brain tumor is a type of disease that damages the brain
through an uncontrolled growth of cells. The details of the
brain tumor is obtained through MRI. For giving right
treatment the analysis of the tumor must be performed
accurately. Segmentation method is used for the purpose of
analysis, and is done to distinguish the brain tumor tissue
from other tissues such as fat, edema and normal tissue. The
MRI image must be maintained at the edge of the first image
with median filtering, followedbysegmentationprocessthat
requires thresholding. Segmentation process is performed
by giving a mark on the area of the brain and area outside
the brain using watershed method then clearing the skull
with cropping. 14 brain tumor images areusedasaninputin
this study. The segmentation result compares brain tumor

area with brain tissue area. The tumor was determined with
average error rate of 10 percent.
2. PROPOSED SYSTEM
As per literature survey, it was found that automated brain
tumor detection is very necessaryashighaccuracyisneeded
when human life is involved. Automated detection of tumor
in MR images involves feature extraction and classification
using machine learning algorithm. In this paper, a system to
automatically detect tumor in MR images is proposed as
shown in figure.
Fig.1:- System Architecture
METHODOLOGY (CNN):
The process took place in two step i.e. training and testing
phase. Training phase always takes place before testing
phase. The feature extraction and classification is done by
convolution neural network (CNN). Training image set are
used to train the model and testing dataset are used to
validate the model. Loss function is used to improve the
accuracy of the model. Less the value of loss function more
accurately the prediction is done. Generally, labelled image
set are used to train the model (Fig 1).
Convolutional neural network (CNN, or ConvNet) is a form
deep learning and most commonly applied to analysing
visual imagery. CNNs use a variation of multilayer
perceptron designed to require minimal pre-processing.
They are also known as shift invariant or space invariant
artificial neural networks (SIANN), based on their shared-
weights architecture and translation invariance
characteristics. Convolutional networks were inspired by
biological processes in thattheconnectivitypattern between
neurons resembles the organization of the animal visual
cortex. Individual cortical neurons respondtostimulionlyin
a restricted region of the visual field known as the receptive
field. The receptive fields of different neurons partially
overlap such that they cover the entire visual field.CNNsuse
relatively little pre-processing compared to other image
classification algorithms.Thismeansthatthenetwork learns
the filters that in traditional algorithms were hand-
engineered. This independence from prior knowledge and
human effort in feature design is a major advantage. They
have applications in image and video recognition,
recommender systems, image classification, medical image
analysis, and natural language processing. A CNN consistsof
an input and an output layer, as well as multiple hidden
layers. The hidden layers of a CNN typically consist of
convolutional layers, pooling layers, fully connected layers
and normalization layers.
Fig2. Simple ConvNet
The Convolutional Neural Network in Fig. is similar in
architecture to the original LeNet and classifies an input
image into four categories: dog, cat, boat or bird. There are
four main operations in the ConvNet shown in fig. above:
1. Convolution
2. Non-Linearity (ReLU)
3. Pooling or Sub Sampling
4. Classification (Fully Connected Layer)
An Image is a matrix of pixel values. Essentially, every image
can be represented as a matrix of pixel value Channel is a
conventional term used to refer to a certaincomponentof an
image. An image from a standard digital camera will have
three channels – red, green and blue – you canimaginethose
as three 2d-matrices stacked over each other (one for each
colour), each having pixel values in the range 0 to 255.
The Convolution Step:
ConvNets derivetheirnamefromthe“convolution”operator.
The primary purpose of Convolution in case of a ConvNet is
to extract features from the input image. Convolution
preserves the spatial relationshipbetween pixelsbylearning
image features using small squares of input data. We will not
go into the mathematical detailsofConvolutionhere, but will
try to understand how it works over images as we discussed
above, every image can be considered as a matrix of pixel
values. Consider a 5 x 5 image whose pixel values are only 0
and 1 (note that for a grayscale image, pixel values range
from 0 to 255, the green matrix below is a special casewhere
pixel values are only 0 and 1.

Also, consider another 3 x 3 matrix as shown. Then, the
Convolution of the 5 x 5 image and the 3 x 3 matrix can be
computed as shown in the animation in Fig below:
Fig3. The Convolution operation
The output matrix is called Convolved Feature or Feature
Map. Take a moment to understand how the computation
above is being done. We slide the orange matrix over our
original image (green) by 1 pixel (also called ‘stride’)andfor
every position, we compute element wise multiplication
(between the two matrices) and add the multiplication
outputs to get the final integer which forms a single element
of the output matrix (pink)[8]. Note that the 3×3 matrix
“sees” only a part of the input image in each stride. In CNN
terminology, the 3×3 matrix is called a ‘filter‘ or ‘kernel’ or
‘feature detector’ and the matrix formed by sliding the filter
over the image and computing the dot product is called the
‘Convolved Feature’ or ‘Activation Map’ orthe‘FeatureMap‘.
It is important to note that filters act as feature detectors
from the original input image.
Introducing Non-Linearity (ReLU):
An additional operation called ReLU has been used after
every Convolution operation in Figure above. ReLU stands
for Rectified Linear Unit and is a non-linear operation. Its
output is given by:
Fig4:- ReLu function
ReLU is an element wise operation (applied per pixel) and
replaces all negative pixel values in the feature map by zero.
The purpose of ReLU is to introduce non-linearity in our
ConvNet, since most of the real-world data we would want
our ConvNet to learn would be non-linear (Convolution is a
linear operation – element wise matrix multiplication and
addition, so we account for non-linearity by introducing a
non-linear function like ReLU).
The Pooling Step:
Spatial Pooling (also called subsampling or down sampling)
reduces the dimensionality of each feature map but retains
the most important information. Spatial Pooling can be of
different types: Max, Average, Sum etc.
In case of Max Pooling, we define a spatial neighborhood(for
example, a 2×2 window) and take the largest element from
the rectified feature map within that window. Instead of
taking the largest element we could also take the average
(Average Pooling) or sum of all elements in that window. In
practice, Max Pooling has been shown to work better.
shows an example of Max Pooling operation on a Rectified
Feature map (obtained after convolution + ReLU operation)
by using a 2×2 window.
Fig5. Max Pooling
We slide our 2 x 2 window by 2 cells (also called ‘stride’)and
take the maximum value in each region. As shown in Figure,
this reduces the dimensionality of our feature map.
3. CONCLUSION
The In summary, we propose a CNN-based method for
segmentation of brain tumors in MRI images. There are
several existing of techniques are available for brain tumor
segmentation and classification to detect the brain tumor.
There are many techniques available presents a study of
existing techniques for brain tumor detection and their
advantages and limitations. To overcome these limitations,
propose a Convolution Neural Network (CNN) based
classifier. CNN based classifier used to compare the trained
and test data, from this get the best result.

ACKNOWLEDGEMENT
We wish to express my profound thanks to all whohelped us
directly or indirectly in making this paper. Finally, we wish
to thank to all our friends and well-wishers who supported
us in completing this paper successfully. We are especially
grateful to our guide Prof. S.R.JADHAV for his time to time,
very much needed, and valuable guidance. Without the full
support and cheerful encouragement of my guide, the paper
would not have been completed on time.
REFERENCES
[1] Nilesh Bhaskarrao Bahadure, Arun Kumar Ray and Har
Pal Thethi ,” Image Analysis for MRI Based Brain Tumor
Detection and Feature ExtractionUsingBiologicallyInspired
BWT and SVM”, Hindawi International Journal ofBiomedical
Imaging volume 2017.
[2] Andras Jakab, Stefan Bauer et al., “The Multimodal Brain
Tumor Image Segmentation Benchmark (BRATS) “IEEE
TRANSACTIONS ON MEDICAL IMAGING, VOL. 34, NO. 10,
2015.
[3] Israel D. Gebru, Xavier Alameda-Pineda, Florence Forbes
and Radu Horaud, “EM Algorithms for Weighted-Data
Clustering with Application to Audio-Visual Scene Analysis“
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND
MACHINE INTELLIGENCE, VOL. XX, NO. Y, 2016.
[4] Prateek Katiyar, Mathew R. Divine et al., “A Novel
Unsupervised Segmentation Approach Quantifies Tumor
Tissue Populations Using Multiparametric MRI:FirstResults
with Histological Validation” Mol Imaging Biol 19:391Y397
DOI: 10.1007/s11307-016-1009-y , 2016.
[5] Zeynettin Akkus, Alfiia Galimzianova,AssafHoogi,Daniel
L. Rubin and Bradley J. Erickson, “Deep Learning for Brain
MRI Segmentation: State of the Art and Future Directions” J
Digit Imaging DOI 10.1007/s10278-017- 9983-4, 2017.
[6] Anupurba Nandi, “Detection of human brain tumour
using MRI image segmentation and morphological
operators” IEEE International Conference on Computer
Graphics, Vision and Information Security (CGVIS), 2015.
[7] Swapnil R. Telrandhe, Amit Pimpalkar and 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
(WCFTR’16), 2016.
[8] Komal Sharma, Akwinder Kaur and Shruti Gujral, “Brain
Tumor Detection based on Machine Learning Algorithms“
International Journal of Computer Applications (0975 –
8887) Volume 103 – No.1, 2014

IRJET- Brain Tumor Detection using Convolutional Neural Network

Recommended

More Related Content

What's hot (20)

Similar to IRJET- Brain Tumor Detection using Convolutional Neural Network (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Brain Tumor Detection using Convolutional Neural Network