Proceedings of the International Conference on Artificial Intelligence and Smart Systems (ICAIS-2021)

IEEE Xplore Part Number: CFP21OAB-ART; ISBN: 978-1-7281-9537-7

Brain Tumor Detection Analysis Using CNN: A

Review
2021 International Conference on Artificial Intelligence and Smart Systems (ICAIS) | 978-1-7281-9537-7/20/$31.00 ©2021 IEEE | DOI: 10.1109/ICAIS50930.2021.9395920

Sunil Kumar Renu Dhir

M.Tech Scholar Associate Professor
Department of Computer Science and Engineering Department of Computer Science and Engineering
Dr. B.R Ambedkar National Institute of Technology Dr. B.R Ambedkar National Institute of Technology
Jalandhar, India Jalandhar, India
[email protected] [email protected]

Nisha Chaurasia
Assistant Professor
Department Of Information Technology
Dr. B.R Ambedkar National Institute of Technology
Jalandhar,India
[email protected]

Abstract—A Brain Tumor is essentially a malformed cell and damage the healthy cell [1]. In MRI image handling,
growth that can be cancerous and non-cancerous. The tumor digital image processing assumes an essential task. In the
in the Brain is the most dangerous disease and can be diagnosed restorative field, X-ray images are typically used for evaluating
easily and reliably with the help of detection of the tumor
with automated techniques on MRI Images. Several methods of and recognizing the body’s tumor growth. Both children and
efficient diagnosis and segmentation of brain tumors have been adults have a common brain tumor. Tumors cause extreme
suggested by many researchers for effective tumor detection. A brain pressure, which spread across the entire brain region.
review method involving two-stage approaches for 20 research Inside the skull, tumor development takes place and interferes
papers published in the period from 2000 to 2020 has been with normal brain function. There could be a tumor leading
conducted to learn about tumor detection in MRI images. There
has also been a comparative study of various processes. Two key to Cancer, a major cause of death and responsible for about
problems are discovered after an intensive analysis procedure, 13 percent of all deaths worldwide, which is a major cause
Image Restoration and Image Enhancement. Which are solved of death. The cancer incidence rate is dynamic increase at an
and clarified with suitable methodologies. The introduction of alarming rate of the world.
quantitative image analysis resulted in fields such as MRI The National Cancer Institute (NCI) has reported that
Images. On-going area for interest in brain tumors, in particular
focus on White Matter, CSF, GM. The novel method uses the 22,070 new cases of brain cancer and other central nervous
CNN classification technique and has been used to disregard systems will be diagnosed in the US in 2009. The American
the dataset picture algorithm error. The implementation of Brain Tumor Association explains this statics further reported
the suggested model is applied in the Python and TensorFlow that 62,930 the new case of a benign stage of brain tumor
environment. Algorithms and methodologies used to solve specific [7].Fig.1 illustrates the presence of a Brain tumor. There is
research problems were included in the results and along with
their strengths and limitations. This examines the quantitative currently no simple identification of the primary cause of brain
characteristics of brain tumors, such as shape, texture, and signal tumors. However, they’re caused by radiation contamination
intensity, to predict high accuracy with a low error rate and the during MRI, CT scan, and X-rays. The following symptoms
potential for future work in the field. confirm the existence of brain tumors, severe seizure, loss of
Index Terms—Brain Tumor, Convolution Neural Network, Self- motor function, convulsions, neurological disturbances, and
organizing map, Restoration, Enhancement, Computed tomogra-
phy, Magnetic resonance imaging issues with numbness, speech and also hormonal abnormalities
and personality change can also symptoms of brain tumor [3]
[4].
I. I NTRODUCTION
The latest traditional diagnostic method is based on the
The brain is one of the maximum complicated organs inside experience of humans in the perception of a decision in MRI
the human body that works with a very large no of cells. scan and arise the risk of false diagnosis and brain tumor
Brain tumors increase when there is an unregulated division recognition as well. Applying, on the other hand, Digital image
of cells that forms an irregular. The group of a cell will affect processing allows the identification of the tumor easily and
the normal functionality of the behavior of the brain activity reliably. A critical point of the research was medical image

978-1-7281-9537-7/21/$31.00 ©2021 IEEE 1061

Authorized licensed use limited to: R V College of Engineering. Downloaded on April 21,2022 at 14:38:39 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Artificial Intelligence and Smart Systems (ICAIS-2021)
IEEE Xplore Part Number: CFP21OAB-ART; ISBN: 978-1-7281-9537-7

brain tumors using image processing has been in use for

many decades. Researchers have indicated Many of the semi-
automatic techniques for detecting brain tumors and many
automatic image processing techniques, but most of them do
not include efficient and reliable outcomes due to the existence
of noise and weak contrasting images that typically occur
throughout the Medical Images [6]. Because of the complex
brain structure and brain tumor segmentation is difficult. But
early and reliable tumor detection is difficult. The diagnostic
Fig. 1. The Presence of Brain Tumor [7] device is very critical for tumors, edema, and necrotic tissues.
Tumors are vulnerable to damage to normal brain cells and
Inflammation is created and the pressure is placed on parts
segmentation, as it inherited complicated issues for the ac- of the brain region and pressure within the skull is increased.
curate segmentation of images in brain disorder. Radiologists Existing currently, the algorithms used in the processing of
use a CT scan and MRI to visually examine the patient. They medical images used partial differential equations, inspired
examine the brain structures, tumor size, and tumor location curate flows, and diverse mathematical trends [5].
was illustrated by MRI Images [10]. The structure of this paper is organized as follows: firstly
Imaging plays a crucial role in Brain tumor diagnosis. They brief introduction about Brain Tumor. Section II is related
often tendon MRI to be either hypo tense or iso tense. As work and Section III is Methodology and Table 1 is Sum-
the images are called edges, immediate shifts in grey tones marization of Classification Accuracy Obtained by Various
occur. Techniques for edge detection convert images to Edge Methods and Section IV is Brain Tumor Data sets and Section
images benefit from the gray tone shifts in the images. The V is Conclusion and future scope.
segmentation image is obtained without any modifications in
the primary image’s physical characteristics. The MRI images II. RELATED WORK
from the radiologists presented details such as the location of
tumors, a simple way to diagnose the tumor, and prepare the In this section, we discussed many existing brain tumor
surgical procedure [2]. detection techniques in the field. S, et al. [3] Describes an
approach for the order of MRI images, that depends on the
back-propagation of neural system procedure. The strategy is
built using the techniques of image enrichment, segmentation,
registration, character recognition, and segregation. During
the segmentation procedure, the morphological operations and
threshold values are considered. These training image and
experiment is analyzed by a neural network technique of back-
propagation algorithm to the recognition of the presence of a
tumor.
S, et al. [5] proposed a novel method for automatic cancer
detection using DNN(deep neural network) for expertise detec-
Fig. 2. a)Normal Brain, b) Benign Tumor, c)Malignant Tumor [17] tion of glioblastomas. Its use final layer which implementation
on fast segmentation, which is an order of 24 seconds to 3
Fig.2 illustrates from the brain region of a tumor indicates, minutes for entire region of lungs.
where the infected part of the brain is located. The three- Jyoti, et al. [6] Latest progress in the use of Magnetic
dimensional representations of the secret organ are obtained Resonance Images in biomedical image processing. Imaging
using a heavy field of magnetism and radio waves in MRI. (MRI) allows fast diagnosis and localization of brain tumors.
The MRI technique has the benefit that ionizing radiation We plan to divide brain scans into eight (8) categories, with
is not at risk. The big bonus of MRI scans is the lack of seven(7) categories indicating various types of tumors and
ionizing radiation. By applying image processing techniques, one category for the normal brain. Using the Leave2-Out
the diagnostic capability of the MRI images is improved. cross-validation process, the suggested classification strategy
In medical image processing, brain image is a critical task. is validated.
The creation of medical photographs makes medical diagnosis Joshi, et al. [7] In recent years, research in the medical
possible by exposing the internal mechanisms of the human imaging area has concentrated a great deal on the segmentation
body’s unseen organs. The processing of medical photographs of brain tumors. Only the quantitative measurement of disease
lowers the burden on both patients and physicians. It is one modeling makes it possible to monitor and recover when doing
of the main optical image processing applications in which so. During the early stages, MR normally. In detecting defects
mathematical operations are applied to an image to increase of the brain, and cerebral infarction, brain tumor, or infection
its accuracy [17]. This method used for the identification of in e, the disease is more susceptible.

978-1-7281-9537-7/21/$31.00 ©2021 IEEE 1062

Authorized licensed use limited to: R V College of Engineering. Downloaded on April 21,2022 at 14:38:39 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Artificial Intelligence and Smart Systems (ICAIS-2021)
IEEE Xplore Part Number: CFP21OAB-ART; ISBN: 978-1-7281-9537-7

Kiranmayee, et al. [8] Proposed an approach to detect the convolutional neural network-based brain tumor classification
brain tumor consisting of training and testing level. Building system. The preparation and assessment stages of the CNN-
a blueprint application has validated the functionality of the based brain tumor classification are separated. The number of
proposed algorithm. The results of the developed prototype photographs is separated into various groups by labeling them
show that the option of emotionally supportive networks in with terms like a tumor and non-tumor brain image, and so
the field of medicinal services can be combined to increase on. Preprocessing, feature extraction, and classification are all
the quality of services. used in the training phase [17].
Arya, et al. [9] presented a review on various image pre-
processing and segmentation techniques like image filtration
method, Denoising method, Histogram based segmentation,
Watershed segmentation, SVM based segmentation, and MRF
based segmentation can be a module for a better result for
accuracy and a lower rate of error.
Demiharan, et al. [14] proposed segmentation methods to
classify MRI in brain tumor. Cerebrospinal fluid (CSF), edema,
white matter (WM), and gray matter are using station wavelet
transform, learning vector quantization, and Average similarity
was found in order of 0.87 of gray matter 0.96 of CSf.0.77 of
edema and 0.91 percent in white matter.
Aneja, et al. [16] proposed the fuzzy clustering means
method and segmentation algorithm, that work with FCM clus-
ter against noise image. Segmentation value is evaluated based
on the cluster validity functions, runtime, and convergence
error rate achieved by 0.537 percent.
Udhaya, et al. [17] Such techniques are investigated using
data mining algorithms, such as apriori, k-means clustering,
k-nearest neighbor classifier, fp tree-based association mining,
and decision tree creation performed an inside and outside
assessment techniques. This experiment completed in this
paper is accompanied by the following techniques: such as
various techniques for data mining algorithm, three methods
are separate. i) In almost all situations, the overhead of the
interface is below 10 percent. ii) For each of the four data
mining algorithms, strong parallel performance is achieved.
iii) For phenomenal execution, the combination of different
strategic swings is necessary if the decision tree construction
algorithm occurs.
Teshnehlab. et al. [35] Convolutional Neural Network
(CNN) was used to detect a tumor. The CNN was the first
place where images were seen. The Softmax Fully Connected
plate, which was used to classify images, had a classification
accuracy of 98.67 percent. The CNN’s precision is also 97.34
percent with the Radial Basis Function (RBF) classifier and
94.24 percent with the Decision Tree (DT) classifier.
III. METHODOLOGY
The human brain is modeled through the design and ap-
Fig. 3. Overall flow of the brain tumor detction system
plication of a neural network. These Papers confer the Brain
Tumor Detection based on brain region photos on MRI Images
using CNN Model. Brain regions are in the first level, extracted A. Image Preprocessing
from the MRI image and any slice in that area is segmented Fig. 4 illustrates the preprocessing of the image. The fol-
to get tumors. The tumor regions that are segmented are lowing preprocessing steps were applied to every image: 1.
used of CNN Architecture. CNN is used to assess the Patient Crop brain containing a section of the images. 2. Transform
Images. The primary aim of this research is to detect Brain the images to the shape of (240, 240,3) as images have come
Tumor. If the tumor present in the brain part of a patient is from different sources. so they may have different images since
malignant or benign. Figure 3 shows a block diagram of a images in the dataset come in different sizes. So all images

978-1-7281-9537-7/21/$31.00 ©2021 IEEE 1063

Authorized licensed use limited to: R V College of Engineering. Downloaded on April 21,2022 at 14:38:39 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Artificial Intelligence and Smart Systems (ICAIS-2021)
IEEE Xplore Part Number: CFP21OAB-ART; ISBN: 978-1-7281-9537-7

need to be in the same form for feeding input to the neural F. Feature Optimization
network. 3. Normalization applied: to scale the pixel values to Feature optimization is the combined process of feature se-
the 0-1 range [17]. lection and extraction which play a crucial role in brain image
processing. The starting feature selection process decreases the
dimensionality of the feature sets and take a minimum time
for detecting. Then extract the best sets of feature raw dataset
Genetic Algorithm(GA) is used [17] [12].

G. Classification
Classification means labeling the image as per requirement
in features. Apply on best feature is recognized by the Genetic
Algorithm and such classifier like CNN for result [2] [26] [25].

Fig. 4. Raw Image and Preprocessed Image [17] H. Convolution Neural Network Architecture
In the area of medical image processing, the Neural Network
B. Image Segmentation is commonly used. Many researchers have tried to develop a
model over the years that can more accurately recognize the
It is the mechanism by which an image is split into tumor. To determine the utility of the proposed brain tumor
regions with different shades, textures, brightness, contrast, classification system, training accuracy, validation accuracy,
and gray characteristics level. A digital grayscale picture is and validation loss are measured. We attempted to come up
an input to the operation. Anomalies are the performance of with an example that could correctly identify the tumor from
the operation. The use of segmentation is to have greater data 3D images of the brain MRI. This tumor can be identified by a
than existing in medical photos. Different processes, such as fully-connected neural network, but we adopted CNN for our
Neural Networks, the decision tree, and the algorithm based on model because of parameter sharing and connection sparsity
rules and Bayesian Networks in a segment are used to obtain [17] [5].
desired performance data [17] [25] [26]. There are many other
segmentation methods.
1) Thresholding Method: As the name suggests, voxels
above the threshold are known as belonging to the tumor.
2) Region growing method: Seed voxel is entered into the
segmentation; from this seed, voxels that are identical are
identified as belonging to the tumour.
3) Region growing method: Changes in the density between
the edges of the voxels are seen as the limits of the tumors.
C. Watershed Algorithm
this is a unique segmentation method whereby the voxel
intensities or gradients are represented by a topographical map
similar to those seen in geography. Based on the ‘steepness’
of the map, a boundary is assigned.
Fig. 5. Convolution Neural Network Architecture [28]
D. Feature Selection
Feature Extraction is a method of collecting an image’s Each x (image) input has a (224 height, 224 widths, 1
visual content. The extraction of features is the method of channel) is given to the neural network and it passes to the
portraying the raw image in its reduced form to promote layers that follow: Fig. 5 illustrates the Working of Neural
decision-making, such as the classification of patterns. After Network such as: a) A layer of 0 fillings and (5, 5) pool size.
Segmentation of the brain, DWT is used for segmentation of i. 32 filter convolution sheet and (8, 8) filter size and stride of
MRI Images [13]. The most important characteristics of the 2. b) A batch normalization layer to normalize values of pixels
chained filter of low pass and high pass to derive features. for computation acceleration. i An activation layer, ReLU. ii
A Max Pooling Layer with f=4 and s=4. iii Same as before,
E. Image Classification a layer of Max Pooling layer with f=4 and s=4. iv A sheet to
Classification means marking the pictures according to level.
their characteristics. The best function is defined through the The CNN is organized in layers and has a hierarchical
application of GA, among other items. The other approach is structure. CNN also has input and output layers, as well as
advanced by involving the GA in three performance compar- convolutional, pooling, normalization, and Fully Connected
ative classifiers, such as CNN and ML [12]. layers. In terms of the number of layers used, the scale and

978-1-7281-9537-7/21/$31.00 ©2021 IEEE 1064

Authorized licensed use limited to: R V College of Engineering. Downloaded on April 21,2022 at 14:38:39 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Artificial Intelligence and Smart Systems (ICAIS-2021)
IEEE Xplore Part Number: CFP21OAB-ART; ISBN: 978-1-7281-9537-7

TABLE I
SUMMARIZATION OF CLASSIFICATION ACCURACY OBTAINED BY VARIOUS METHODS

Author Methods Data Set Accuracy Merits Demerits

Wentao et al. [1] Deep CNN and BraTS2014 and SVM 87.05% and CNN Segmentation and less time Algorithm is not
SVM BraTS2016 86.69% optimized
Deepak, et al. [3] Deep CNN and Figshare SVM classifier 0.97% and Improved performance and robust- Poor
GoogLeNet DCNN 92.3% ness performance
of transfer model
Jyoti et al. [6] Deep CNN and OASIS 93.18%accuracy, 94% Significantly improvement for The gradient is
SVM precision, 93% recall and multi-class classification vanishingly small
92% f1-score and consequently
prevent
Yang et al. [13] Discrete wavelet GE Healthcare Clustering accuracy of Instead of dimensionality reduction Model fitting is-
transform (DWT) 94.8% and a balanced on SVM sue arise
error rate of 7.8%
Demirahan et al. Wavelets, Neural IBSR2015 and WM 91%, GM87%, Performance increases in WM, It should be
[14] Networks and BRATS2012 edema 77%, tumor 61% GM, CSF and edema implemented on
self-organizing and CSF 96% most updated
map (SOM) data sets as well
Aneja et al. [16] Fuzzy Clustering NSL-KDD FCM 1.283, T2FCM Reduce the noise from the training Least percentage
Means 0.962 and IFCM 0.537 set and size clot of misclassifica-
Algorithms tion error
S et al. [19] Fuzzy logic, K BRATS2010 FCM in WM, GM, CSF Improve noisy image Low error Highly reduced
means and Neu- 29.60, 30.50 and 29.60 rate misclassification
ral Networks error rate
Badza et al. [21] CNN BRATS and 10-fold cross-validation Applied on two different data Set. High running
CBICA accuracy was 96.56% time
Seetha et al. [27] CNN, KNN and BRATS 2015 CNN perform better with Complexity and computation time Training time is
SVM an accuracy of 97.5% is low and accuracy is high high

number of images used, and the form of activation functions vector that can be used as an input for the next level. The
used, CNN is special and using trial and error as a basis for production of both convolution and pooling layers constitutes
experimentation [34]. a significant portion of the input image’s characteristics. For
1) Convolution Layer: CNN’s (convolutional neural net- our system, we used a Convolution Neural Network to train
works) are a form of neural network that is commonly used our model on a variety of datasets [17] [18].
for image recognition and classification.CNN is divided into
three categories. Le Net, Alex Net, and Google Net are their IV. B RAIN T UMOR DATASET
names. CNN is mostly used to classify images. We provide
the image as input, but the machine converts it to a pixel Tensorflow and Keras, developing a Brain tumor detection
array. To collect the function map, a filter is slid over the model using a CNN. The dataset combination of Br35H
input grayscale graphic. Another function map would result 2020 and Central research UK and the dataset includes three
from the convolution of another filter on the same image. Local categories: yes, no and prediction file containing 3060 MRI
dependencies can be determined from the given original image images of the brain, and yes contain. 1500 MRI images of
using the convolution operation. CNN can get filthy during the tumorous and no contains 1500 brain MRI images of non-
preparation period [33]. tumorous MRI images and prediction contain 60 brain MRI
images [29] [30] [31].
2) ReLU: Rectified Linear Unit (ReLU) is a nonlinear
phase abbreviation. Both negative pixel values are changed
to 0. ReLU is also implemented for the non-linearity equation A. Brain Tumor Cases
on CNN. As we take data from the real world for this model, Fig. 6 illustrates the Cancer Research UK, brain, other CNS
we use ReLU. So, we’re Using CNN to read the details [33] intracranial tumor incidence related to age, with the highest
[32]. incidence relate increasing in older people. In central research
3) Pooling(downsampling): Pooling is a dimensionality re- UK in 2015-2017. On Avg, no cases increase in each year
ductional algorithm that reduces each function map’s dimen- 23 percent in older age in 75 and most types of brain and
sion. There are many forms of pooling. Max pooling, Average Intracranial tumors also increasing rapidly in younger age.
pooling, and Sum pooling are the three options. The net value This maximum reflects cell DNA damage from time to time.
is replaced by the nearby outputs in the pooling function. Normal MRI does not see flowing fluid, such as blood in an
Max Pooling takes the largest element from the window and artery, and this produces flow voids that appear in the picture
modifies all the nearby elements in that window, replacing as black holes. Mostly detected a tumor in brain recognized
them with the large element [18] [33]. in a malignant stage, because a small tumor is not detected in
4) Fully connected Layer(flatten): It takes the previous MRI, CT scan and X-ray images [29]. Fig. 7 illustrates the
layers’ output, ”flattens” it, and transforms it into a single no of brain tumor case in india [30].

978-1-7281-9537-7/21/$31.00 ©2021 IEEE 1065

Authorized licensed use limited to: R V College of Engineering. Downloaded on April 21,2022 at 14:38:39 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Artificial Intelligence and Smart Systems (ICAIS-2021)
IEEE Xplore Part Number: CFP21OAB-ART; ISBN: 978-1-7281-9537-7

R EFERENCES
[1] Wu, Wentao, et al. ”An Intelligent Diagnosis Method of Brain MRI Tu-
mor Segmentation Using Deep Convolutional Neural Network and SVM
Algorithm.” Computational and Mathematical Methods in Medicine
2020 (2020).
[2] Nguyen, Phong Thanh, et al. ”Brain tumor segmentation and classifica-
tion using KNN algorithm.” (2019).
[3] Deepak, S., and P. M. Ameer. ”Brain tumor classification using deep
CNN features via transfer learning.” Computers in biology and medicine
111 (2019): 103345.
[4] Abd-Ellah, Mahmoud Khaled, et al. ”A review on brain tumor diagnosis
from MRI images: Practical implications, key achievements, and lessons
learned.” Magnetic resonance imaging 61 (2019): 300-318.
[5] Sasikala, S., M. Bharathi, and B. R. Sowmiya. ”Lung Cancer Detection
and Classification Using Deep CNN.”, International Journal of Innova-
tive Technology and Exploring Engineering (IJITEE) 8.2S (2018).
[6] Islam, Jyoti, and Yanqing Zhang. ”A novel deep learning based multi-
class classification method for Alzheimer’s disease detection using brain
MRI data.” International conference on brain informatics. Springer,
Fig. 6. Average Number of New Cases per Year and Age-Specific Incidence Cham, 2017.
[29]
. [7] Joshi, Deepika, and Rohit Goyal. ”Review of Tumor Detection in Brain
MRI Images.” (2017).
[8] Kiranmayee, B. V., T. V. Rajinikanth, and S. Nagini. ”Explorative
Data Analytics of Brain Tumour Data Using R.” 2017 International
Conference on Current Trends in Computer, Electrical, Electronics and
Communication (CTCEEC). IEEE, 2017.
[9] Arya, Madhvi, and Reecha Sharma. ”Brain Tumor Detection through
MR Images: A Review of Segmentation Techniques.” International
Journal of Computer Applications 975 (2016): 8887.
[10] Ali, Eman M., Ahmed F. Seddik, and Mohamed H. Haggag. ”Automatic
detection and classification of Alzheimer’s disease from MRI using
TANNN.” International Journal of Computer Applications 148.9 (2016).
[11] Veer, Shubhangi S., and Pradeep M. Patil. ”Brain tumor classification
using artificial neural network on MRI images.”,International Journal of
Research in Engineering and Technology 4.12: 218-226(2015).
[12] Zhao, Feng, Hanqiang Liu, and Jiulun Fan. ”A multiobjective spatial
fuzzy clustering algorithm for image segmentation.” Applied Soft Com-
puting 30 (2015): 48-57.
[13] Yang, Guang, et al. ”Discrete wavelet transform-based whole-spectral
Fig. 7. No of Brain Tumor Surgery in India [30] and subspectral analysis for improved brain tumor clustering using single
voxel MR spectroscopy.” IEEE Transactions on Biomedical Engineering
62.12 (2015): 2860-2866.
[14] Demirhan, Ayşe, Mustafa Törü, and İnan Güler. ”Segmentation of tumor
V. CONCLUSION AND FUTURE SCOPE and edema along with healthy tissues of brain using wavelets and
neural networks.” IEEE journal of biomedical and health informatics
19.4 (2014): 1451-1458.
In this paper existing method counter many challenges [15] Balakumar, B., P. Raviraj, and E. Divya Devi. ”Brain Tumor Classifi-
such as accuracy, quality of tumor, and tumor detection time. cation Using Machine Learning Algorithms.”, Elysium Journal P-ISSN:
In this paper, we are indicated at the various method of 2347-4408 (2017).
[16] Aneja, Deepali, and Tarun Kumar Rawat. ”Fuzzy clustering algorithms
an advanced novel brain tumor identification method. For for effective medical image segmentation.” International Journal of
preprocessing MRI images, the original preprocessed segment Intelligent Systems and Applications 5.11 (2013): 55-61.
applies median filtering techniques and validation accuracy [17] Kunam, A. Udhaya, TamilNadu Sivagangai, and India Dr N. Sujatha.
”Brain Tumor Segmentation Based On a Various Classification Algo-
is 92 percent. Such acquired characteristics were given as rithm”.Volume 4, Issue 12, December(2016).
feedback for three thousand. These classifier performances [18] Guo, Lei, et al. ”Tumor detection in MR images using one-class immune
according to the term of sensitive emotion, accuracy, and feature weighted SVMs.” IEEE Transactions on Magnetics 47.10 (2011):
validation testing. CNN methods are best for the accuracy level 3849-3852..
[19] Kumar, S. Sathish, et al. ”An improved method of segmentation using
with a lower rate of error. So the target region is segmented fuzzy-neuro logic.” 2010 Second International Conference on Computer
and the determination of the existence of the tumor using the Research and Development. IEEE, 2010.
technique proposed here lets doctors make the treatment plan [20] Maulik, Ujjwal. ”Medical image segmentation using genetic algorithms.”
IEEE Transactions on information technology in biomedicine 13.2
and condition of the tumor surveillance in the diagnosis. The (2009): 166-173.
benefits of this method are it increases the segmentation level [21] Badža, Milica M., and Marko Č. Barjaktarović. ”Classification of brain
and spatial localization of the image and thus improves the tumors from MRI images using a convolutional neural network.” Applied
Sciences 10.6 (2020): 1999.
performance relative to the other system. It takes less time to [22] Hall, Lawrence O., et al. ”A comparison of neural network and fuzzy
compute and is faster to train than other networks with fewer clustering techniques in segmenting magnetic resonance images of the
parameters. The accuracy method is most arise using CNN. In brain.” IEEE transactions on neural networks 3.5 (1992): 672-682.
[23] Shen, Shan, et al. ”MRI fuzzy segmentation of brain tissue using neigh-
this future work like improving the accuracy with a low rate borhood attraction with neural-network optimization.” IEEE transactions
of error using different classifier techniques. on information technology in biomedicine 9.3 (2005): 459-467.

978-1-7281-9537-7/21/$31.00 ©2021 IEEE 1066

Authorized licensed use limited to: R V College of Engineering. Downloaded on April 21,2022 at 14:38:39 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the International Conference on Artificial Intelligence and Smart Systems (ICAIS-2021)
IEEE Xplore Part Number: CFP21OAB-ART; ISBN: 978-1-7281-9537-7

[24] SivaSankari, S. M. S. R. S., et al. ”Feature extraction of brain tumor

using MRI.” International Journal of Innovative Research in Science,
Engineering and Technology 3.3 : 10281-10286 (2014).
[25] AM, Athul Sukumar, and Peter Augustine. ”Efficient brain tumor
classification using PCA and SVM.” International Journal of Research
in Engineering, IT and Social Sciences 7 (2017): 1-7.
[26] Badisa, Harshini, Madhavi Polireddy, and Aslam Mohammed. ”CNN
Based brain tumor detection.” International Journal of Engineering and
Advanced Technology (IJEAT) 8.4 (2019).
[27] Suhara, S. Fathima, and M. Safish Mary. ”Fully Connected Pyramid
Pooling Network (FCPPN)–A Method For Brain Tumor Segmentation.”
11.3 : 1457 (2018).
[28] Mohamed Ali Habib:”Brain Tumor Detection”,
’https://github.com/MohamedAliHabib/Brain-Tumor-Detection,(2019)
[29] Cancer Research Institute UK:”Cancer Statistics UK, 2015-
2017”,’https://www.cancerresearchuk.org/health-professional/cancer-
statistics/statistics-by-cancer-type/brain-other-cns-and-intracranial-
tumours/incidence’,(2017).
[30] Neuro Conditions: ”Low Cost Brain Tumor Surgery in
India”,’https://www.spineandneurosurgeryhospitalindia.com/neuro-
conditions/low-cost-brain-tumor-surgery-in-india’,(2017).
[31] https://www.kaggle.com/ahmedhamada0/brain-tumor-detection,(Br35H:
Brain Tumor Detection,2020)
[32] Rai, Hari Mohan, et al. ”A Novel Deep CNN Model for Classification of
Brain Tumor from MR Images.” 2020 IEEE 1st International Conference
for Convergence in Engineering (ICCE). IEEE, 2020.
[33] Seetha, J., and S. Selvakumar Raja. ”Brain tumor classification using
convolutional neural networks.” Biomedical and Pharmacology Journal
11.3 (2018): 1457.
[34] Aghdam, Hamed Habibi, and Elnaz Jahani Heravi. ”Guide to convolu-
tional neural networks.” New York, NY: Springer 10 (2017): 978-973.
[35] M. Siar and M. Teshnehlab, ”Brain Tumor Detection Using Deep Neural
Network and Machine Learning Algorithm,” 2019 9th International Con-
ference on Computer and Knowledge Engineering (ICCKE), Mashhad,
Iran, 2019, pp. 363-368, doi: 10.1109/ICCKE48569.2019.8964846.

978-1-7281-9537-7/21/$31.00 ©2021 IEEE 1067

Authorized licensed use limited to: R V College of Engineering. Downloaded on April 21,2022 at 14:38:39 UTC from IEEE Xplore. Restrictions apply.