BRAIN TUMOR DETECTION USING CNN & ML TECHNIQUES

© 2023, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 654
BRAIN TUMOR DETECTION USING CNN & ML TECHNIQUES
Mary Mathalin A#1, Karthiga T*2
1,2Department of Electronics Communication and Engineering,
Karpaga Vinayaga College of Engineering and Technology,
Chengalpattu-603308, India
---------------------------------------------------------------------------***----------------------------------------------------------------------
Abstract
The human brain is the main controller of the
humanoid system. Abnormal growth and division of
brain cells causes brain tumor, and the proliferation of
brain tumor causes a brain tumor. In the field of human
health, computer vision plays an important role because
it reduces human judgment and provides accurate
results. CT scans, X-rays and MRIs are the most common
imaging methods among the most reliable and safe
magnetic resonance imaging (MRI) scans. MRI detects
objects every minute. The purpose of our article is to
focus on the use of different techniques to detect brain
cancer through brain MRI. In this study, we performed
preprocessing using a bilateral filter (BF) to remove
noise from MRI images. This was followed by binary
thresholding and convolutional neural network (CNN)
segmentation techniques for reliable detection of tumor
regions. Training, testing and validation datasets are
used. With our machine, we predict whether the subject
has a brain tumor or not. The results obtained are
tested using various proven performance parameters
including accuracy, sensitivity and specificity. It is
assumed that the performance of the proposed work is
better compared to similar ones.
Introduction
Medical image processing refers to a variety of
possible technologies. It is used as a non-invasive
method to examine the inside of the body. Medicine
Imaging includes various imaging methods and
processes Imaging the human body for treatment and
diagnosis thus plays the most important and essential
role in activities for the benefit of society. Improves
people's health. Image segmentation is an essential and
essential step of images. Processing that determines the
success of a higher level image Processing .
The main purpose of image segmentation in
medicine Image processing is mainly used to detect
tumor and lesions and is effective. Perform image
processing and obtain satisfactory results for further
diagnosis. Increased sensitivity and specificity for tumor
or lesions With the help of computers, this is becoming
a central issue in medical imaging. Supported diagnostic
systems (CAD). Besides them, there are also other
cancers of the brain and nervous system. 10th cause of
death and 5-year survival Brain tumor account for 3 %
of men and 36% of women. In addition, the World
Health Organization (WHO) states: 00,000 people
worldwide have a brain tumor In recent years, 120,000
people have died. In addition, Anne Arviot has 86,970
new cases of primary malignant and non-malignant
tumor in the brain and other central nervous system
and# 0;CNSand# 1; tumor It is expected to be diagnosed
in the United States in 2019. Brain tumor occur when
abnormal cells form in the brain. There are mainly two
types of tumor: malignant tumor and benign tumor.
Malignant brain tumor start in the brain, grow rapidly,
Actively penetrates the surrounding tissues. Can be
contagious to others Part of the brain affects the central
nervous system. Cancers can be classified as primary
tumor that start from a tumor. Brain tumor or
secondary tumor that have spread from elsewhere are
So-called metastatic tumor of the brain. On the other
hand, benign brain tumor are groups of cells that grow
relatively slowly in the brain. Therefore, early detection
of brain tumor can play an important role. A role in
improving treatment options and maximizing benefits
Survival options are within reach. However, manual
segmentation Removing tumor and lesions is a time-
consuming, complex and difficult task because many of
his MRI images are generated as a medical routine. MRI,
also called magnetic resonance imaging It is mainly
used to detect brain tumor and lesions. brain tumor
Segmentation of magnetic resonance imaging is one of
the most important tasks in medicine Image processing
(usually requiring a lot of effort) data.
In addition, the tumor may be poorly confined to
the soft tissues. limit Therefore, obtaining accurate
results is a very labor-intensive task Segmenting tumor
from the human brain. In this article, we have suggested
an effective and clever method. It helps segment and
identify brain tumor based on both traditional
classifications without human assistance. and
convolutional neural networks.
Literature Survey
One of the most difficult and demanding tasks is
the segmentation of the region of interest of the object,
one of them is the segmentation of tumor from MRI
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 09 | Sep 2023 www.irjet.net p-ISSN: 2395-0072

brain images. ambitious Scientists around the world are
investigating. Use this field for the most segmented ROI
and different variations. A simulated approach from a
specific perspective. recently Neural network-based
segmentation gives excellent results The results and
application process of this model are as follows.
Growing every day.
Devkota et al. [7] Based general segmentation
Processing based on mathematical morphological
operations Spatial FCM algorithm to improve
calculations We have time, but the proposed solution
has not been tested yet. Evaluation phase and results:
Detected cancer 92% Classifier accuracy is 86.6%. Yntao
et al. [8. rank] This was similar to the histogram-based
segmentation method. We consider the brain tumor
segmentation task as three classes of tasks. (tumor
including necrosis and tumor, edema and normal tumor
Classification problem with two categories Flare and T1.
Abnormal region was detected as follows. Region-based
active contour model of FLAIR model. of Outlier, edema
and tumor tissue regions were extracted based on
contrast K- method and Dice coefficients for T1
modality are generated and Sensitivity was 73.6% and
90.3%.
Based on an edge detection approach, Badran et
al. [9] A jointly developed intelligent edge detection
model Adaptive thresholding method for ROI extraction
is adopted. dataset It contained 102 images. The image
was then processed first. The two sets of neural
networks used Canny Edge detection from the first set,
which was matched to the second set. A threshold
calculation is carried out. The segmented image looks
like this: Represented with layer number and functional
properties Extracted using the Harris method. Then we
have two neural networks that were originally used to
detect healthy people and brains containing tumor, and
the other is used to detect tumor. Presentation of
results and comparison between the two Canny Edge
detection methods showed a better performance model
in accuracy results.
Pei et al. [10] Technology that uses tumor growth
patterns as a new method Ability to improve tumor
structure-based segmentation Longitudinal MRI.
Marker maps are used for tumor search Modeling of
proliferation and prediction of cell density after
extraction Structure (such as fractals and mBm) and
intensity features. The performance of the model is
reflected in the average DSC. Tumor cell density - LOO:
0.819302 and 3 folders: 0.82122. Dina et al. We
implemented a model based on [11]. Stochastic neural
network models for vector quantization learning. The
model was evaluated using 6 MRI images, using 18 MRI
images as the test set and the rest as the training set.
Gaussian filter The image is now smoother. The
processing time was reduced by 79% with the modified
PNN method. stochastic neural network, Othman et al.
applied segmentation technique. Principal component
analysis (PCA) was used for function. Unpack and
reduce large dimensions as well.
Data [12]. MRI images are converted into matrices,
where a probabilistic neural network is used for
classification. Finally, a performance analysis is
performed. data of training It contained 20 subjects and
test data of its 15 subjects. Topic Based on the value of
the spread, the accuracy is: 73% - 100%. A region-based
fuzzy clustering and deformable model,
Rajendran et al. [13] achieved 95.3.1% of ASM and
Jaccard index based on a stochastic fuzzy C-means
model with multiple morphological elements.
Zahra et al. [1 ] Works with LinkNet network for
tumor segmentation. Initially, a single link network was
used, which was connected to the network and sent all 7
training datasets to this network segmentation. I did not
consider the angle We present a method that can
automatically process images and CNN segment the
most common brain tumor where this occurs. No
preprocessing steps are required. Achieve a dice score
of 0.73 0.79 on a single network and 0.79 on multiple
systems.
Proposed Methodology
Our proposed methodology has two different
models For segmentation and detection of brain tumor.
In first model Tumor were segmented according to FCM
and classified according to conventional classification.
and his second model tumor were detected using deep
learning..
System Architecture
MRI
Imazzzzzzzzzzzzz
Skull Stripping
Pre-processing
Morphological
Operation
Segmentation
using FCM:

FIG: flow diagram for brain tumor detection
Skull Stripping: :Skull stripping is a very important
step in medical image processing because of the
background of the MRI image not containing any useful
information, and it only increases the processing time.
In our work, we removed the skull portion from the MRI
images in two. These two steps are:
a) Otsu Thresholding: b) Connected Component
Analysis.
Pre-processing :For better segmentation, we need to
maximize the MRI image quality with minimized noise
as brain MRI images are more sensitive to noise than
any other medical image. Gaussian blur filter was used
in our work for Gaussian noise reduction existing in
Brain MRI which prevailed the performance of the
segmentation.
Segmentation using FCM:The Fuzzy C-Means
clustering algorithm was employed for segmentation,
enabling data to be assigned to multiple clusters. This
resulted in the generation of a fuzzy clustered
segmented image, thereby ensuring improved
segmentation accuracy.
Morphological Operation: In order to isolate the
tumor, it was necessary to separate the brain region
from the skull region. To achieve this, morphological
operations were applied to the images. Initially, erosion
was performed to separate weakly connected regions
within the MRI image. Subsequently, dilation was
applied to further refine the segmentation.
Tumor Contouring: Tumor contouring was
accomplished through an intensity-based approach
known as thresholding. This method facilitated the
extraction of the tumor cluster, resulting in an image
where the tumor area is highlighted against a dark
background.
Feature Extraction: For classification purposes, two
types of features were extracted from the segmented
MRI images. Texture-based features, including
Dissimilarity, Homogeneity, Energy, Correlation, ASM,
as well as statistical-based features such as Mean,
Entropy, Centroid, Standard Deviation, Skewness, and
Kurtosis, were utilized.
Traditional Classifiers: To evaluate the accuracy of
tumor detection in our proposed model, we employed
six traditional machine learning classifiers: K-Nearest
Neighbor, Logistic Regression, Multilayer Perceptron,
Naïve Bayes, Random Forest, and Support Vector
Machine.
Evaluation Stage: In the evaluation stage, we compared
our proposed segmentation technique with other
region-based segmentation methods. Our model
demonstrated superior accuracy in segmenting the
region of interest (ROI) and effectively segregating the
tumor portion.
Proposed Methodology Using CNN
The Convolutional Neural Network (CNN) is
widely utilized in the field of Medical image processing.
Numerous researchers have endeavoured to construct a
model that can efficiently detect tumor. Our objective
was to develop an exemplary model capable of
accurately classifying tumor from 2D Brain MRI images.
While a fully-connected neural network can detect
tumor, we opted for CNN due to its parameter sharing
and sparsity of connection.
For tumor detection, we have introduced and
implemented a Five-Layer Convolutional Neural
Network. This aggregated model, comprising seven
stages including hidden layers, yields the most
prominent results in apprehending tumor.
To begin, the convolutional layer is employed as
the initial layer, generating an input shape for the MRI
images of 64*64*3, thereby converting all the images
into a homogeneous dimension. Subsequently, by
consolidating all the images in the same aspect, a
convolutional kernel is created and convoluted with the
input layer. This convolutional kernel consists of 32
convolutional filters, each with a size of 3*3, supported
by 3 channels tensors. The Rectified Linear Unit (ReLU)
is utilized as the activation function to ensure
compatibility with the output.
In this Convolutional Neural Network (ConvNet)
architecture, the spatial size of the depiction is
Tumor
Contouring
Feature Extaction
Traditional
Classifiers
Evaluation Stage

progressively reduced in order to decrease the number
of parameters and computational time required by the
network. When working with Brain MRI images, there is
a risk of overfitting, which can be mitigated by using a
Max Pooling layer. For spatial data that supports our
input image, w horizontally e utilize the MaxPooling 2D
layer in the model. This convolutional layer operates on
a dimension of 31*31*32. Due to the division of the
input images in both spatial dimensions, the pool size is
set to (2, 2), indicating a down sampling by a factor of
two vertically and .Following the pooling layer, a
pooled feature map is obtained. Flattening is a crucial
step after pooling, as it involves transforming the entire
matrix representing the input images into a single
column vector. This step is imperative for further
processing. The flattened vector is then fed into the
Neural Network for further processing.
Two fully connected layers were employed Dense-1
and Dense-2 represented the dense layer. The dense
function is applied in Keras for the processing of the
Neural Network, and the obtained vector is work as an
input for this layer. There are 128 nodes in the hidden
layer. Because the number of dimension or nodes
proportional with the computing resources we need to
fit our model we kept it as moderate as possible and for
this perspective 128 nodes gives the most substantial
result. ReLU is used as the activation function because
of showing better convergence performance. After the
first dense layer, the second fully connected layer was
used as the final layer of the model. In this layer, we
used sigmoid function as activation function where the
total nu mber of the node is one because we need to
lower the uses of computing resources so that a more
significant amount assuages the execution time. Though
there is a chance of hampering the learning in deep
networks for using of the sigmoid as the activation
function, we scale the sigmoid function, and the number
of the nodes is much lesser and easy to handle for this
deep network
Experimental Results
We used six classifiers algorithm and proposed
model for detaction.To find performance evaluation we
used following methods.
After using the algorithms we’ve got accuracies with
special fashions:
Classifiers Accuracy (%)
K-Nearest Neighbor 89.39
Logistic Regression 87.88
Multilayer Perception 89.39
Naïve Bayes 78.79
Random Forest 89.39
PROPOSED 92.42
We achieved an accuracy of 92.42%, which is quite
remarkable when considering the utilization of a five-
layer CNN.
Conclusion
We proposed a computerized method for the
segmentation and identification of a brain tumor using
the Convolution Neural Network. The input MR images
are read from the local device using the file path and
converted into grayscale images. These images are pre-
processe for the elimination of noises that are present
inside the original image. and Convolution Neural
Network segmentation is applied, which helps in
figuring out the tumor region in the MR images. The
proposed model had obtained an accuracy of 92.47%
and yields promising results without any errors and
much less computational time.
Fig: Confusion-matrix

REFERENCES
[1] Kasban, Hany & El-bendary, Mohsen & Salama, Dina.
(2015). “A Comparative Study of Medical Imaging
Techniques”. International Journal of Information
Science and Intelligent System. 4. 37-58.J. Clerk
Maxwell, A Treatise on Electricity and Magnetism, 3rd
ed., vol. 2. Oxford: Clarendon, 1892, pp.68–73.
[2] D. Surya Prabha and J. Satheesh Kumar,
“Performance Evaluation of Image Segmentation using
Objective Methods”, Indian Journal of Science and
Technology, Vol 9(8), February 2016.
[3] Brain Tumor: Statistics, Cancer.Net Editorial Board,
11/2017 (Accessed on 17th January 2019)
[4] Kavitha Angamuthu Rajasekaran and Chellamuthu
Chinna Gounder, Advanced Brain Tumour Segmentation
from MRI Images, 2018
[5] General Information About Adult Brain Tumor". NCI.
14 April 2014. Archived from the original on 5 July
2014. Retrieved 8 June 2014. (Accessed on 11th January
2019)
[6] M. Young, The Technical Writer’s Handbook. Mill
Valley, CA: University Science, 1989.
[7] B. Devkota, Abeer Alsadoon, P.W.C. Prasad, A. K.
Singh, A. Elchouemi, “Image Segmentation for Early
Stage Brain Tumor Detection using Mathematical
Morphological Reconstruction,” 6th International
Conference on Smart Computing and Communications,
ICSCC 2017, 7-8 December 2017, Kurukshetra, India.
[8] Song, Yantao & Ji, Zexuan & Sun, Quansen & Yuhui,
Zheng. (2016). “A Novel Brain Tumor Segmentation
from Multi-Modality MRI via A Level-Set-Based Model”.
Journal of Signal Processing Systems. 87.
10.1007/s11265-016-1188-4.
[9] Ehab F. Badran, Esraa Galal Mahmoud, Nadder
Hamdy, “An Algorithm for Detecting Brain Tumor in
MRI Images”, 7th International Conference on Cloud
Computing, Data Science & Engineering - Confluence,
2017.
[10] Pei L, Reza SMS, Li W, Davatzikos C, Iftekharuddin
KM. “Improved brain tumor segmentation by utilizing
tumor growth model in longitudinal brain MRI”. Proc
SPIE Int Soc Opt Eng. 2017.
[11] Dina Aboul Dahab, Samy S. A. Ghoniemy, Gamal M.
Selim, “Automated Brain Tumor Detection and
Identification using Image Processing and Probabilistic
Neural Network Techniques”, IJIPVC, Vol. 1, No. 2, pp. 1-
8, 2012.
[12] Mohd Fauzi Othman, Mohd Ariffanan and Mohd
Basri, “Probabilistic Neural Network for Brain Tumor
Classification”, 2nd International Conference on
Intelligent Systems, Modelling and Simulation, 2011.
[13] A. Rajendran, R. Dhanasekaran, “Fuzzy Clustering
and Deformable Model for Tumor Segmentation on MRI
Brain Image: A Combined Approach,” International
Conference on Communication Technology and System
Design 2011.
[14] Sobhaninia, Zahra & Rezaei, Safiyeh & Noroozi,
Alireza & Ahmadi, Mehdi & Zarrabi, Hamidreza &
Karimi, Nader & Emami, Ali & Samavi, Shadrokh.
(2018). “Brain Tumor Segmentation Using Deep
Learning by Type Specific Sorting of Images”.

BRAIN TUMOR DETECTION USING CNN & ML TECHNIQUES

More Related Content

Similar to BRAIN TUMOR DETECTION USING CNN & ML TECHNIQUES

More from IRJET Journal

Recently uploaded

BRAIN TUMOR DETECTION USING CNN & ML TECHNIQUES