Brain Tumor Detection and Classification Using Int

This article has been accepted for publication in IEEE Access.
This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3242666
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.Doi Number
Brain Tumor Detection and Classification using

Intelligence Techniques: An Overview
Shubhangi Solanki1, Uday Pratap Singh2, Siddharth Singh Chouhan3, and Sanjeev Jain4
1
Department of Computer Science Engineering, LNCT University, Bhopal, 462042, India
2
School of Mathematics, Shri Mata Vaishno Devi University, Kakrayl, Katra, 182320, India
3
School of Computing Science and Engineering, VIT Bhopal University, Bhopal-Indore Highway, Kothrikalan, Sehore, Madhya Pradesh, 466114, India
4
Department of Computer Science and Information Technology, Central University of Jammu, 181143, India
Corresponding authors: Shubhangi Solanki (e-mail: solanki.shubhangi@gmail.com) and Siddharth Singh Chouhan
(siddharth.smvdu@gmail.com).
ABSTRACT A tumor is carried on by rapid and uncontrolled cell growth in the brain. If it is not treated in
the initial phases, it could prove fatal. Despite numerous significant efforts and encouraging outcomes,
accurate segmentation and classification continue to be a challenge. Detection of brain tumors is significantly
complicated by the distinctions in tumor position, structure, and proportions. The main disinterest of this
study stays to offer investigators, comprehensive literature on Magnetic Resonance (MR) imaging's ability
to identify brain tumors. Using computational intelligence and statistical image processing techniques, this
research paper proposed several ways to detect brain cancer and tumors. This study also shows an assessment
matrix for a specific system using particular systems and dataset types. This paper also explains the
morphology of brain tumors, accessible data sets, augmentation methods, component extraction, and
categorization among Deep Learning (DL), Transfer Learning (TL), and Machine Learning (ML) models.
Finally, our study compiles all relevant material for the identification of understanding tumors, including
their benefits, drawbacks, advancements, and upcoming trends.
INDEX TERMS Brain Tumor, Image Classification, Image Segmentation, Deep Learning, and Machine
Learning.
I. INTRODUCTION because surgery alone would be insufficient to treat them.

An unchecked expansion of brain tissues is known as a brain Aden squamous astrocytoma is an indication of such a tumor.
tumor. It produces pressure in the skull and interferes with the Grade IV: The most dangerous and likely to spread malignant
brain's natural functioning. Brain tumor comes in two different tumors are in this category. They might even use blood vessels
types: Benign (non-cancerous) and Malignant (cancerous). to speed up their growth. An illustration of one of these tumors
Among them, malignant tumors grow quickly in the brain, is glioblastoma multiforme [3]-[5].
damage the normal tissues, and may replicate themselves in Brain tumors must be identified in time and appropriately
other parts of the body [1]-[3]. Brain tumors are graded into be classified in order to get proper treatment and endure for
four different categories: patients. Because of the several vulnerabilities including
Grade I: These tumors do not spread quickly and develop different shapes, sizes of tumors, appearance, positions,
slowly. These are connected to a higher chance of enhanced scanning parameters, and modalities detection of brain tumors
order and may be surgically eliminated nearly entirely. One is a very challenging job to perform [5]. To attain this task a
such tumor is a pilocytic astrocytoma. number of traditional and intelligence techniques are being
Grade II: Although they may migrate to surrounding tissues used. Typically, traditional approaches like Leksell Gamma
and advance to higher grades, these tumors also grow over Knife®, Gamma Knife (GK), and Radioactive beams are
time. These tumors may detect even though treatment is taken helpful in diagnosing the lesions, but this process includes
by the patient. An oligodendroglioma tumor is an example of human involvement and is often a time-consuming task to
an overtime growth tumor. perform [6]. For brain tumor identification, many medical
Grade III: The growth of these tumors has been quicker than imaging modalities like Computer Tomography (CT),
grade II malignancies and could spread to adjoining tissues. Magnetic resonance imaging (MRI) scans, and Poisson
Such tumors require post-operative chemo or radiotherapy Emission Tomography (PET) are employed. Also, A unique
VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
Solanki et al.: Survey of Deep Learning approaches used for Brain Tumor
MR technique called chemical exchange saturation transfer machine learning mechanisms adopted for the detection and
(CEST) makes it possible in imaging some substances at classification of brain tumors from MRI images.
concentrations that are too low to affect the contrast of 2)Study was carried out for about 100 articles collected from
conventional MR imaging and too low to be directly identified various sources like ScienceDirect, Springer, IEEE, etc.
in MRS at usual water imaging resolution. Among them, MRI 3)A separate analysis of both approaches has been carried out
scan is a non-invasive method that shows the internal body and various findings are tabularized individually.
structure with the help of magnetization and microwave 4)Further, a research gap analysis has been carried out to
pulses. For brain tumor diagnosis, three categories of magnetic differentiate between the importance of DL over ML.
resonance image patterns are used: Fluid Attenuated Inversion 5)Various findings like datasets, deep models, classification
Recovery (FLAIR), T1 weighted, and T2 weighted. The approaches, parameters, future research directions along with
problem of identifying and detecting tumor-infected areas the importance of using 3D models and attention-based
using brain MRI is crucial [6]. mechanisms are being discussed followed by our proposed
The human visual system has a minimal ability to notice work.
tiny variations brought on by the Magnetic Resonance Image’s Organization of manuscript: introduction in section I is
increased complexity (MRI). Recently, a number of followed by a literature review given in section II. It includes
investigators developed Systems for computer-aided the various studies categorized among DL and ML in a
diagnosis (CAD) to help radiologists make precise diagnoses separate section. A research gap analysis of the underlying
[6]. Although Leksell Gamma Knife® is a better approach to technologies is also been given in this section. Further, the
diagnosing tumors, because of the presence of necrosis in the various findings are given in Section III. Section IV
brain the finding suffers. Therefore, effective machine demonstrates our proposed work. The manuscript ends with a
learning should be adopted in order to solve this problem. conclusion followed by references.
Authors in [7] have proposed a novel method with the
amalgamation of a Random Forest classifier along with a II. LITERATURE REVIEW
voxel clustering algorithm. Similarly, conventional diagnosis A. DEEP LEARNING TECHNIQUES
processes including Leksell Gamma Knife® are time- In recent years, a lot of research has been directed toward the
consuming processes, therefore authors in [8] have introduced adaptation of deep learning models in diagnosing brain
a semiautomated method using an unsupervised FCM tumors. Academicians have put in their efforts and with the
clustering algorithm for accurately segmenting the lesion help of high-end computing devices, higher accuracy has been
volume. A pipeline of four algorithms namely K-means, achieved. Convolutional neural networks (CNN), which
Fuzzy K-means and Gaussian Mixture Model (GMM), and include input, output, hidden layers, and hyperparameters, are
Gaussian Hidden Markov Random Field (GHMRF) has been often called Deep Learning (DL) [5]. It uses supervised
proposed for the segmentation of brain tumors by the authors classification and generates feature maps by having the kernel
in [9]. Authors in [10] propose a two-stage mechanism for the convolve all around the input image. Automatic-based feature
assessment in dose escalation and eliminate the need for extraction is both possible with DL models. Apart from its
multispectral MRI data to analyze the image. The proposed usefulness for medical condition detection, it has some
framework incorporates the FCM algorithm in defining a shortcomings, including the requirements to design complex
novel method named a fully automatic method for necrosis models, fine-tuning of hyper-parameters, the requirement of
extraction (NeXt). Although ML approaches are quite large data set, and time and effort to training/testing. As per
efficient in handling the MRI images for accurate detection of recent research, significant data augmentation methods like
resizing, rotation, scaling, and transformation are enforced to
the tumor region, with the availability of complex, large
tackle the big data availability problem. A trained NN is used
volumes of data and high computing devices, deep learning
in transfer learning techniques to extract similar properties
models are being cast-off for achieving advanced
from an application-specific dataset [1]. For brain tumor
performance. Therefore, to understand the detailed learning identification current TL methods like RESNET-100,
mechanism of these intelligence techniques the proposed work VGGNET, Google-Net, AlexNet, etc. are applied. The various
is aimed in presenting: deep-learning techniques used by the researchers in the past
1)The proposed work incorporates various deep learning and are summarized in Table I.
TABLE I
BRAIN TUMOR DETECTION USING DEEP LEARNING TECHNIQUES
Reference Methodology Algorithms Accuracy Dataset used
Machiraju Jaya Brain tumor classification using Softmax Inception -V3 89.00% 3064 MRI images
Lakshmi et al. classifier
[1]
2 VOLUME XX, 2017
Wen Jun & Brain tumor Classification using CNN 98.61% 3043 MRI images
Zheng Liyuan attention mechanism
[2]
Arshia Rehman Three-layer architecture of CNN with VGG-16 and CNN 98.69 % MRI brain slices of 233 patients
et.al. [3] transfer learning technique
Tharindu Theoretical examination of common Various Deep and machine NA Biomedical images, electrical biomedical
Fernando [4] deep learning algorithms for detecting learning techniques signals, and other biomedical data such
medical anomalies as data from laboratory results, audio
recordings, and wearable medical
devices
Alexander Supervised learning was used across the Quantitative susceptibility 95.20% Various Data sets of brain MRI
Selvikvåg MRI processing pipeline from mapping and CNN
Lundervold et. acquisition to retrieving, segmentation to
al. [5] illness prediction
Yun Jiang et al. Multiscale CNN based statical threshold Convolutional Neural 86.30% MICCAI BRATS2015 dataset
[11] pattern techniques Network technique
Dongnan Liu et Multi-dimensional image data Deep Convolutional Neural 86.50% MICCAI BRATS 17 Challenge
al. [12] processing on network portals Network (DCNN)
Muhammad Deep CNN has been used for the CNN earning Methods NA Positron Emission Tomography (PET)
Waqas et. al. organization images tool is used for assessing brain
[13] tumors and differentiating tumor
progression from reactive changes
Yakub Detection and classification using CNN CNN 77.60% The chosen MR image dataset consists of
Bhanothu et. al. three classes namely glioma,
[14] meningioma, and pituitary tumors.
Zhiguan Huang, Convolution Neural Network has been Deep CNN 95.50% A brain tumor dataset provided by
et. al. [15] used with various optimization and Nanfang Hospital, Guangzhou, China,
different activation functions for the and General Hospital, Tianjing Medical
analysis of complex tumors. University, China, from 2005 to 2010
Nyoman Convolutional Neural Network Deep CNN 84.20% A brain tumor dataset consisting of 3064
Abiwinanda et T-1 weighted CE-MRI images publicly
al. [16]. available via figshare Cheng
Ali Ari and Convolutional Neural Network ELM-LRF 97.20% Digital Imaging and Communications in
Davut Hanbay Medicine (DICOM) file format. cranial
[17] magnetic resonance (MR) images were
classified as benign or malignant by
using ELM-LRF.
Yota Ishikawaet Deep CNN, PNN and PNN Binarization algorithm with 98.50% The number of brain cell images of our
et al. [18] water shade feature extraction dataset is 720, and its ratio of astrocyte to
model low-grade astrocytoma is one-to-one, so
each class has 360 images
Heba Mohsen et Deep Neural Network (DNN) PCA and DWT classification 96.90% Human brain MRIs with 22 normal and
al. [19] 44 abnormal images which are
glioblastoma, sarcoma, and metastatic
bronchogenic carcinoma tumors
collected from Harvard Medical School
Justin S. Paul et Numerous Deep learning techniques and CNN 91.40% brain tumor dataset
al. [20] Machine learning techniques
Yan Xu et al. Deep CNN with various Instigation Various deep learning 97.50% The training data is provided by the
[21] Structures frameworks for features organizers from the TCGA web
extraction and classification
Kaoutar B. Deep Convolutional Neural Networks Fine-Tuning 81.00% Dataset provided by the H. Lee Moffitt
Ahmed et al. (DCNNs) Cancer Research Center
[22]
Mustafa Rashid Deep Neural Networks Classification of tumors using 94.00% Brain MRI slices were collected from
Ismael [23] SoftMax and sparse Nanfung Hospital in Guangzhou, and the
autoencoder General Hospital of Tianjing Medical
University in China during the period
from 2005 to 2010
Renhao Liu et DCNN Detail feature classification of 95.40% MRI dataset. From top to bottom, they
al. [24] brain tumor Magnetic are TI-weighted (post gadolinium), T2-
Resonance
2 VOLUME XX, 2017
weighted, FLAIR, and ADM from the H.

Lee Moffitt Cancer Research Center.
Nøhr Ladefoged CNN RESOLUTE and DeepUTE 67.105 NA
et al. [25]
Himar Fabelo et 2D CNN Multi-layer CNN 80.30% NA
al. [26]
Yuexiang Li & CNN Multi-view DNN 88.00% BraTS 17 dataset is utilized.The dataset
Linlin Shen [27] contains multimodal MRI scans of
glioblastoma (GBM/HGG) and lower-
grade glioma (LGG).
Lina Chato & Normal discriminant analysis SVM, KNN, LR, and 68.80% The glioma brain dataset was provided
Shahram Latifi different learning algorithms. by the BraTS 2017 Challenge
[28]
Virupakshappa Artificial Neural Network (ANN) Modified Level Set approach 98.00% The BRATS database contains four
& Basavaraj sequences of MRI images for every
Amarapur [29] patient which includes FLAIR, T1, T1
with Contrast, and T2 weighted images.
The BRATS 2015 database.
Eze Benson et CNN Heterogeneous feature for 92.00% The dataset of BraTS 2018
al. [30] classification
Chenhong Zhou CNN U-Nets were deepened by 81.35% Dataset of BraTS 2018 challenge
et al. [31] adding double convolution
layers, emergence modules,
and dense modules
Geena Kim [32] 2D Fully CNN CNN Architecture 88.20% BRATS15 and BRATS17 dataset
Parnian Afshar Feature extraction with DCNN Capsule Networks (Caps 86.56% Brain Magnetic Resonance Imaging
et al. [33] Nets) (MRI) image
Peter D. Chang Feed Forward Neural Network Fully connected CNN 87.40% BRATS 2016 challenge
[34]
Fabian Isensee DCNN U-Net Architecture 90.10% BraTS challenge
et al. [35]
Sanjay Kumar et Deep-CNN UNET Architecture 89.60% Brats Dec 2017 dataset (a)T (b) T1
al. [36] contrast-enhanced (c) T3 (d) flair
AM. Hasan et.al A deep learning approach to feature Hybrid deep learning features 99.30% MRI axial slices from Iraqi center for
[37] withdrawal may be used to excerpt Collaborative deep learning research and magnetic resonance of Al-
information from MRI brain images. To features are used Kadhimain Medical Cit
extract customized features, use
MGLCM
S. Deepak et.al Deep learning approaches of training Deep CNN-SVM 97.10% Figshare dataset and brain MRI images
[38] models for feature extraction
Hari Mohan Rai U-net model and VGG-16 CNN 98.00% MRI dataset with cropped and uncropped
& Kalyan images
Chatterjee [39]
Deepak V.K & Brain tumor Segmentation using deep Fully CNN - BRATS 2018 dataset
Sarath R [40] learning with cascaded regression
technique
Ejaz Ul Haq Data augmentation technique CNN 97.30%, Figshare and BRATS 2018 dataset
et.al. [41] 96.50%
Sidra Sajid et.al. Deep learning CNN 86.00% BRATS 2013 dataset
[42]
Ramin Brain tumor segmentation using deep Cascade CNN 92.03% BRATS 2018 dataset
Ranjbarzadeh learning and Attention-based mechanism
et. Al [43]
T. Ruba et.al Brain tumor segmentation using novel Cascaded CNN,3D U-net - BRATS 2015
[44] LSIS operator from 3D MRI images
Aman Verma et Brain tumor classification using deep CNN 97.87% MRI images
al. [45] model
2 VOLUME XX, 2017
Praveen Kumar Brain tumor detection using CNN CNN 98.9%5 MRI images
Ramtekkar et. al
[46]
R. Rajasree et al. Brain tumor classification using deep U-NET, CNN 96.36% BRATS 2015
[47] learning
With the recent developments in technology, 3D scanning is tumors implemented with a 3D medical image. In this context,
also being used for the analysis of tumors. 3D image classification is performed using two distinct classifiers:
processing for brain tumor detection and classification has Dense-Net and Dark-Net. On the BRATS 2018 dataset of 3D-
been described in [48]. It used various deep learning MR images, the suggested framework obtained an accuracy of
frameworks, such as MobileNetV2, MobileNetV3 small, 98.67% and a dice similarity coefficient (DSC) of 97.91% for
MobileNetV3 big, VGG16, VGG19, and custom CNN segmentation. For brain tumor classification on 3D-MR
models. CNN achieved the highest accuracy. It offers a images from the BRATS 2018 dataset, the suggested
solution that combines a CNN built with Keras and Tensor framework obtained a DSC of 98.14%, an accuracy of 98.26%
flow with a fully-featured cross-platform application built using the Dense-Net classifier, and a DSC of 96.4%, an
with PyQt5 and MariaDB, all of which are designed for usage accuracy of 96.52% using the Dark-Net classifier. A higher
in medical settings like hospitals and clinical images. The level of accuracy was achieved by the Dense-Net classifier
primary goal of this work is to characterize a brain damaged compared to the Dark-Net classifier. In addition, they have
by a tumor using real-world data and identify abnormal pixels compared this framework to earlier research, and the results
[51]. In [52] authors describe a pre-processing, data show that custom CNN obtains higher classification accuracy.
augmentation, segmentation, and binary classification of brain Some of the 3D-based methods are given be Table II.
TABLE II
BRAIN TUMOR DETECTION USING 3D DEEP LEARNING TECHNIQUES
Md. Akram Hossan Tuhin 3D MRI, MRSI, and CT images are used for the CNN and 3D CNN 85.005 Real-time hospital image
et. al. [48] detection of brain tumors using CNN and the dataset
detection of the tumor using segmentation
methods
Rudresh D. Shirwaikar et. 3D CNN-based brain tumor segmentation, tumor Various machine - Some real-time and some
al. [49] detection, and classification learning and deep synthetic datasets are utilized
learning classifiers.
H. Yahyaoui et. al. [50] DensNET using custom CNN for brain tumor 3D CNN , 92.06%, 2 Heterogenous dataset
85.00%
detection DENSENET model including 3D and 2D real-time
images
Pokhrel S et. al. [51] 3D image processing for brain tumor detection MobileNetV2, 92.00 %- Ream word brain image
and classification MobileNetV3 small, 95.00% dataset
MobileNetV3 big,
VGG16, VGG19,
Gull S and Akbar S [52] Binary classification of brain tumors has been Dense-Net and Dark- 96.52% BRATS 2018
implemented with 3D medical image Net based CNN
Yannick Suter et al. [53] 3 Dimensional CNNs SVM with 72.20% BraTS 2018
heterogeneous
feature extraction in
CNN classification
Yan Hu & Yong Xia [54] DCNN 3D-based deep CNN 81.40% BraTS 2017 Challenge dataset
Dong Nie et al. [55] Classification using 3D deep Convolutional 3D multi-level 99.60% Brain images (i.e., T1 MRI,
Neural Network representation and fMRI and DTI) of high-grade
organization by using glioma patients.
CNN
Anand kumar & P.V. Brain tumor segmentation using 3D deep 3DCNN 99.8% BRATS 2015
Shridevi [56] learning
2 VOLUME XX, 2017
Girija Chetty et.al. [57] Medical image analysis by using 3D deep 3D U-NET - BraTS Challenge 2018 dataset
learning
Zeeshan Shaukat. et al. Semantic segmentation using 3D deep learning 3D U-NET Dice score-95% BRATS dataset
[58]
Hiba Mzoughi. et al. [59] Brain Tumor Classification using 3D multiscale. 3D CNN 96.49 % BRATS 2018 dataset
Joseph Stember & Classification of 3D MRI brain tumors Deep Reinforcement 100% testing set 3D MRI images
Hrithwik Shalu [60] learning accuracy
Yohan Jun et al. [61] Detection of metastatic brain tumors using deep 3D CNN 97.08% and Clinical data.
learning 3D black blood technique sensitivity-100%
Pranjal Agrawal.et.al. Segmentation and classification using 3D deep 3D U-net, Deep CNN 90% MRI Kaggle dataset
[62] learning
Agus Subhan Akbar.et.al U-net architecture for brain tumor segmentation Residual attention 77.73%, BRATS 2018,2019,2020,2021
82.19%, 89.33%
[63] block mechanism challenge database
B. MACHINE LEARNING METHODS

Matter (GM), Cerebrospinal Fluid (CSF) as well as White
Pre-processing, segmentation, extraction of features, and Matter (WM). Segmentation techniques that draw inspiration
categorization are the four key phases of ML techniques used from nature include Particle Swarm Optimization (PSO) and
to diagnose brain tumors. Genetic Algorithm. Recent studies show that DL frameworks
1. Preparation: To produce accurate diagnoses in the clinical like Convolutional Neural Networks (CNN), Mask- Recurrent
field, precise imaging is essential. The efficiency of clinical Neural Networks, and UNET outperform conventional
images is influenced by the artifact acquisition methods, like methods in segmentation.
magnetic resonance scans, CT, and PET. A Magnetic 3. Feature Extraction: While extracting features, properties
resonance scan's real images could contain a lot of unwanted of brain MR scans such as shape, structure, wavelet, and
and pointless details. Magnetic resonance imaging is impacted Gabor are retrieved. The Gray-Level Co-occurrence Matrix
by Rician noise [36]. It is challenging to remove Rician (GLCM) is commonly studied. A second-order statistical
distortion since it is signal-sensitive. Pre-processing method is used to evaluate textural features like energy,
techniques including filtration, intensity correction, and skull correlation, and intensity. Wavelet data is derived using the
stripping is being used to maintain the original visual Discrete Wavelet Transform (DWT). The approximation
characteristics. coefficients are obtained and it is applied to an original image,
2. Segmentation: It is a technique used to obtain areas of and then the feature vector is selected. Both automatic features
interest from digital images. The tumor's position must be produced by DL techniques like Convolutional Neural
distinguished from the MR brain scans, which is crucial. For Networks, ResNet, Capsule Networks, and handwritten
segmentation, numerous supervised methods are available, features have shown success. To decrease the number of
including thresholding, soft computing technique, atlas-based, features, PCA and Genetic Algorithms are utilized.
Neural Networks (NNs), clustering, etc. Thresholding 4. Classification: Benign and malignant tumors are the most
methods include global, adaptable, Otsu's, and histogram- prevalent forms of brain tumors. The three types of malignant
dependent techniques. There are two unsupervised clustering tumors include hypothalamic, gliomas, and malignant tumors.
methods namely K-means clustering and fuzzy C-means Table III shows a summary of some ML methods.
clustering. It successfully separates brain MRI scans into Gray
TABLE III
MACHINE LEARNING METHODOLOGIES FOR BRAIN TUMOR DETECTION
HT. Zaw et.al Naive Bayes classification successfully Naive Bayes (NB) 94.00% REMBRANDT database for brain
[64] identifies a tumor location with all classification cancer imaging contains MRI images
propagating malignant tissues
E. Sert et.al Extracted features, as well as classification, The maximum entropy 95.00% Brain tumors in MRI images
[65] are done using the training to the model calculation method has
architecture and support vector machine, been used
correspondingly
TL. Narayana Multi-objective genetic heuristic optimizing Segmentation, Feature 91.23% Datasets provides different MRI
et al. [66] and SVM on brain MRI images Extraction, and SVM images such as T1-W, T2-W, PD, and
Classification FLAIR
2 VOLUME XX, 2017
FP. Polly et.al The computerized system employs k-means Clustering, Segmentation, 99.00% DATABASE OF HGG AND LGG
[67] for clustering used for feature extraction with Feature Extraction and TUMOR.
principal component analysis and discrete Reduction, and SVM
wavelength transform Classification
J. Amin et.al A computerized brain magnetic resonance SVM Classification 98.00% datasets such as Harvard, RIDER and
[68] approach may distinguish malignant from Local.
noncancerous lesions
N. Gupta et.al Non-invasive and adaptable tumors Support vector machine 98.90% Dataset collected from MP MRI &
[69] identification approach using T2-weighted (SVM) classifies CT Scan Centre at NSCB Medical
brain MR images. Entropy measures two College, Jabalpur
important textural and form aspects from the
segmented picture. SVM classifies MR
images using key properties
N. Gupta et.al Naive Bayes-based decision support system Naive Bayes 97.83% Glioma tumors images collected
[70] detects and grades tumor severity from NSCB Medical College
Jabalpur, India, and BRATS dataset
A.Minz et.al Preprocessing, feature extraction, and GLCM features ML and 89.90% brain MR images.
[71] classification using CNN DL-based algorithms
AS. Shankar Gustafson-Kessel fuzzy clustering. Gray Gustafson-Kessel (G-K) 95.00% NA
et.al [72] Level Co-occurrence Matrix extracts MRI fuzzy clustering
image features (GLCM)
C. GAP ANALYSIS OF ML AND DL METHODS incorporating both technologies which are presented in Table
Further, to analyze the research gap between the existing IV. This table compromises the details with respect to
machine learning and deep learning approaches, the study has methodology, algorithms, gap analysis, and dataset used by
been directed towards summarizing the various literature work the authors.
TABLE IV
GAP ANALYSIS OF MACHINE LEARNING AND DEEP LEARNING METHODS
Reference Methodology Algorithms Gap Analysis Dataset used
Tharindu Fernando Theoretical examination of Various ML and DL It is a theoretical evaluation only PhysioNet/CinC 2016 dataset
[4] common DL methods for methods
detecting medical anomalies.
Alexander Selvikvag Supervised learning was used Quantitative susceptibility Various heterogeneous features Image synthesis in MRI
Lundervolda et al. across the MRI processing mapping and CNN are considered for model training
[5] pipeline from acquisition to which generates redundancy
retrieving, segmentation to issues.
illness prediction
Yun Jiang et al. The Multiscale Convolutional Deep CNN High communication cost for MICCAI BRATS 2015 dataset
[11] Neural Networks multiple hidden layers
(MSCNN) with Statistical
Thresholding
Dongnan Liu et al. CBICA's Image Processing Deep CNN It is applicable for the CBICA MICCAI BRATS 17 Challenge
[12] Portal uses a 3D Large Kernel dataset only not work for real-time
Anisotropic Network. datasets.
Muhammad Waqas Deep CNN has been used for the Deep Learning Methods Conventional pretrained CNN Brain tumor
Nadeem, classification modules are used.
Mohammed A. Al
[13]
Yakub Bhanothu, For the detection and CNN with various VGG Generate high-time computation MR image dataset
Anandha narayanan categorization of diseases in modules such as VGG16 when it deals with 50 and 100-
Kamala kannan, MRI scans, CNN has been and VGG32 deep layers.
Govindaraj employed.
Rajamanickam [14]
Zhiguan Huang et al. With a Modified Activation modified CNNBCN, Generates good results for the A brain tumor dataset provided
[15] Function, a CNN Based on ResNet, DenseNet and RESNET-101 module only. by Nanfang Hospital,
Complicated Networks for BT MobileNet Guangzhou, China, and General
classification Hospital, Tianjing Medical
University, China, from 2005 to
2010
Nyoman CNN using deep learning Convolutional Neural Different feature extraction a brain tumor dataset T-1
Abiwinanda et al. method Networks (CNNs) come in methods are used that select non- weighted CE-MRI images
[16] several shapes and
2 VOLUME XX, 2017
dimensions. VGG16, essential and redundant features publicly available via figshare
ResNet, AlexNet sometimes Cheng
Ali Ari & Davut Deep CNN for classification ELM-LRF Based Extreme Low accuracy for all experiments images are in Digital Imaging
Hanbay [17] Learning Machine and Communications in
Medicine (DICOM) file format.
cranial magnetic resonance
(MR) images were classified as
benign or malignant by using
ELM-LRF.
Heba Mohsen et al. Deep Neural Network Principal Components Low accuracy for both PCA and human brain MRIs images from
[19] Analysis (PCA) and DWT Harvard Medical School
Discrete Wavelet
Transform (DWT)
Yan Xu et al. [21] Features of Deep Convolutional Deep Convolutional ImageNET deep framework has The training data is provided by
Activation Activation Features were been used to generate overfitting the organizers from the TCGA
trained using the issues with different optimizers. web
information from Datasets
Kaoutar B. Ahmed et CNN using the deep learning Fine-Tuning High error rate Dataset provided by the H. Lee
al. [22] methodology Moffitt Cancer Research Center
Mustafa Rashid Deep-NNs Softmax with the Stacked High-time computation for deep brain MRI images and BRATS
Ismael [23] Sparse Autoencoder (SSA) layers dataset
Nøhr Ladefoged et CNN RESOLUTE and DeepUTE Low accuracy and dice score -
al. [25]
Himar Fabelo et al. 2D-CNN Multi-layer CNN High computation when a large -
[26] number of conventional layers are
used
Yuexiang Li & CNN SPNet and the Multi-view Time consuming process that BraTS 17 dataset is utilized.The
Linlin Shen [27] Deep Learning Framework generates similar results as CNN dataset contains multimodal
(MvNet) MRI scans of glioblastoma
(GBM/HGG) and lower-grade
glioma (LGG).
Lina Chato & Deep learning was used to Machine learning Conventional machine learning The glioma brain dataset was
Shahram Latifi [28] create the CNN and Linear algorithms include the methods are used provided by the BraTS 2017
Discriminant (LD). support vector machine, LD Challenge
Virupakshappa & Adaptive-ANN Modified Level Set Low accuracy and overfitting The BRATS database contains
Basavaraj Amarapur approach problem four sequences of MRI images
[29] for every patient which includes
FLAIR, T1, T1 with Contrast
and T2 weighted images. The
BRATS 2015 database.
Eze Benson et al. Deep CNN module Pyramid form is used in The dataset of BraTS 2018
[30] Hourglass structure
Chenhong Zhou et Deep learning-based CNN From a range of angles, Both CNNs are conventional Dataset of BraTS 2018
al. [31] OM-Net is used to enhance modules challenge
performance
Geena Kim [32] 2D Fully CNNs Double convolution layers, Fully CNN may take high BRATS15 and BRATS17
dense, and inception computation, it required HPC and dataset
modules were added to a U- GPU environments.
Net to create a deep
architecture
Parnian Afshar et al. CNNs- based deep learning Capsule Networks Low accuracy when evaluated brain Magnetic Resonance
[33] (CapsNets) with heterogeneous datasets. Imaging (MRI) image
Peter D. Chang [34] Fully CNNs Semantic segmentation Feed Forward model has been BRATS 2016 challenge
with a fully convolutional used it generates similar accuracy
neural network (FCNN) as ANN
Fabian Isensee et al. Convolutional Neural Networks UNET is semantic Segmentation features are used for BraTS challenge
[35] segmentation architecture. module training that may not
match generating the background
knowledge
Sanjay Kumar et al. Fully CNNs Deep CNN with different High computation when epoch Brats Dec 2017 dataset (a)T (b)
[36] optimizers size is high T1 contrast-enhanced (c) T3 (d)
flair
Yan Hu & Yong Xia Deep learning-based DCCN Cascaded U-Net 3D DNN- DNN generates low accuracy over BraTS 2017 Challenge dataset
[54] based Technique CNN and DCNN
Dong Nie et al. [55] DL-based 3D CNNs. 3-dimensional multi- High-time computation for brain images (i.e., T1 MRI,
channel architecture CNNs channel mapping and feature fMRI, and DTI) of high-grade
for data transfer and selection in each channel glioma patients.
Support Vector Machine
2 VOLUME XX, 2017
Subhashis Banerjee In addition to classification, Deep Convolutional Neural Different optimizers varied the BraTs 2017 dataset
et.al. [74] CNN has been used for a variety Networks (ConvNets) result for accurate detection of
of feature extractions normal as well as anomaly
Asma Naseer et. al. A Convolution Neural Network CNN and CAD are deep Overfitting problems generate a benchmark dataset, BR35H,
[75] (CNN)-based early brain tumor learning-based algorithms. when it deals with cross-model containing brain tumor MRIs.
detection using six distinct detection six different datasets, i.e., BMI-
datasets. I, BTI, BMI-II, BTS, BMI-III,
and BD-BT.
Isselmou Abd El Deep CNN has used 400 hidden Deep wavelet Auto- Homogeneous encoder features MR brain images from
Kader et.al. [76] layers and numerous SoftMax Encoder model are extracted only not considering BRATS2012, BRATS2013,
functions other features BRATS2014, BRATS2015,
2015 challenge, and ISLES
Lilly Sheeba et. al. By segmenting grayscale Segmentation using Otsu Otsu methodology works on MRI brain images.
[77] images, the Otsu approach has homogeneous features only
been used to locate brain
tumors.
Javaria Amin et. al. A comprehensive study for For brain tumor analysis, It is an overview-based approach Datasets for brain tumor
[78] brain tumors using MRI, image enhancement and theoretical results are detection
including anatomy of brain techniques, segmentation, discussed; no real implementation
tumors and publicly accessible extraction of features, is there
datasets classification, DL &
transfer learning, and
quantum ML are used.
Mohd Shahajad, SVM feature extraction for SVM Old conventional SVM has been MRI images from the Kaggle
Deepak et al. [79] brain tumor MRI image used that not able to extract deep dataset
categorization features
Esther Alberts et al. SVM, MLP, RF, KNN, and LBP, BRIEF, and HOG - a brain tumor dataset from “The
[80] PCA based on ML Cancer Imaging Archive”
(TCIA)
Iqbal, S et al. [81] Segmentation and clustering Several visual Old features extraction and brain tumor multi-classification
characteristics are retrieved selection methods are used
and examined
A. R. Deepa & W. R. Backpropagation algorithm Combined Feature High error rate and low accuracy BRATS 2015 dataset
Sam Emmanuel [82] with Firefly optimization Classification using an for the backpropagation model
adaptive firefly
backpropagation neural
network
Javeria Aminrt et al. Random forest (RF) classifier Gabor Wavelet Many old feature extraction and BRATS 2012, 2013 and 2014
[83] using ML techniques Characteristics, Local segmentation methods are datasets.
Binary Patterns, some of introduced that produce low
the features include accuracy.
segmentation-based fractal
texture analysis and scatter
plot of oriented gradients.
McKinley et al. [84] CNN Dense net is a semantic In dense layer data dropout 2017 BRATS challenge
segmentation algorithm problem
that replaces pooling with
dilated convolutions
Aparna Natarajan et. Resourceful Separation of Brain CNN PNN and DNN Low accuracy when evaluated MRI images dataset
al. [85] Tumor Using fast learning with default optimizers
technique
Ben naceur, M. et. Recognition of brain growths DCNN (Dense-MultiOCM Conventional preprocessing and BRATS-2018 dataset
al. [86]0 using deep approach-based ) noise removal techniques are used
CNN for real-time data
Amiri, S. et. al. [87] Structure RF and Bayesian RF, ANN, and CNN The Cross trained module does not Brain Tumor Image
network for brain tumor work for the heterogamous Segmentation Challenge
classification classifier. (BRATS, 2015) dataset
III. FINDINGS global thresholding. Deep learning-based segmentation allows

Expert radiologists do brain tumor segmentation and for more precise tumor extraction [26]. GLCM and DWT
classification. ML and DL may help radiologists to make largely extract features. GLCM returns texture characteristics,
better decisions. This paper summarizes current strategies for whereas DWT returns approximation coefficients. Deep
automated brain tumor categorization. Histogram learning architectures automate feature extraction. ResNet [4,
equalization, median, Gaussian, and Wiener filters preprocess 11]. PCA and bio-inspired algorithms like PSO are used to
MRI images. There are six forms of segmentation: clustering, reduce dimensionality. Choosing the optimal characteristics
statistical, CNN, region, and threshold-based [37]. K-means for categorization is challenging. Hence a hybrid technique
Researchers often utilize C-means clustering and adaptive integrating several features is utilized. ML and DL techniques
2 VOLUME XX, 2017
are used to classify data. multi-kernel SVM Binary

classification uses linear, RBF, and Cubic. These findings are Accuracy Accuarcy
comparable to VGG19 and ResNet. ANFIS, a fuzzy-ANN 100
hybrid, performs better for binary classification. However, the 98
database does not capture all tumor forms and grades. Or they 96
have to obtain MRIs from nearby hospitals. As a result,
94
comparing the performance of various approaches is difficult.
ACCURACY
A common database of all tumor kinds is required for future 92
study. The deep learning approach can extract more detailed 90

features from the dataset for segmentation and classification. 88
Transfer learning techniques provide better prediction results 86
for deep learning approaches in the effective detection of brain
84
tumors. The machine learning approach gives better
performance when the dataset is small, whereas, with large
datasets deep learning, models are efficient. The deep learning
approach used several pre-processing techniques like scaling REFERENCES
and normalization to enhance desired features. Pre-processing

techniques of machine learning including filtration, intensity FIGURE 1. Evaluation of machine learning techniques for brain tumor
detection.
correction, and skull stripping are being used to maintain the
original visual characteristics with a limited data set. The
primary drawback of machine learning technology is that it is Accuracy Accuracy
100
complicated, with a large number of parameters increasing
90
with the execution time and system requirements for
80
implementation. The deep learning approach offers low
70
complexity, where the features are self-learned by the
ACCURACY
60
network. Figure 1 and Figure 2 demonstrate the valuation of 50
Machine Learning and Deep Learning methods over accuracy. 40
Further, the various findings from the literature are being 30
discussed in separate categories like datasets used, different 20
tumor classification approaches, deep learning models, 10
parameters used, limitations of existing approaches, and future 0
research directions.
A. DATASETS USED
The researchers make use of a variety of datasets that are REFERENCES
available to the general public in order to test the proposed
FIGURE 2. Evaluation of deep learning methods for brain tumor
methodologies. In this section, we will go through various detection.
challenging datasets that are both significant and crucial. The
BRATS datasets are considered to be the most difficult MRI BRATS 2017 dataset [78] was utilized by Shubhashis
datasets [73, 74, 75]. BRATS Challenge is issued at different Banerjee and Francesco Masulli. This dataset comprises data
times throughout the years, and more recent challenges have from the BRATS 2012/13/14, and 2015. A total of 210 HGG
had a resolution of 1 mm3 voxel [76]. Employed two cases and 75 LGG instances of brain tumors are included in
benchmark datasets and one dataset obtained from qualified the dataset. The patient's MRI scan includes four distinct MRI
radiologists. These datasets include 15 photographs of sequences: the initial (T1) sequence, the T1 & T2 weighted
patients, and each patient included 9 slices of imaging data. sequence, and the Fluid Attenuated Inversion Recovery
The core dataset that was used was known as the digital (FLAIR) volume with 155 two-dimensional slices at a
imaging and communication in medicine (DICOM) dataset. resolution of 240 by 240. The BRATS training dataset, which
Twenty-two photos from the DICOM collection, some of consists of 274 multi-modality MRI images of people with
which depict tumor-infected brain tissue, have been taken into gliomas, is used by the researchers Ali Isin, Cem Direkoglu,
consideration for the purpose of this investigation. This dataset and Melike sah (both high and low grades). For the purposes
did not contain any images that represented the ground truth. of testing, a total of 110 scans were taken from ground truths
The brain web dataset [77] was used as a supplemental source and unknown grades.
of information for this study. The whole three-dimensional
B. TUMOR CLASSIFICATION APPROACHES
simulated brain MR data that is included in this paper were
The input data is sorted using classification techniques into a
obtained using three modalities: proton density-weighted
variety of separate classes., after which training and validation
MRI, T1, and T2-weighted MRI, and T1-weighted MRI. The
VOLUME XX, 2022 9
are carried out using both known and unknown instances. The recall of 95.07 percent, and an F1-score of 94.94 percent,
classification of tumors into relevant classifications is a correspondingly. Convolutional neural networks, Deep CNN,
widespread application of machine learning, tumor as well as dual-force CNN, cascaded CNN, 3-dimensional CNN, and
non-tumor, and malignant and benign tumors. Supervised Modern deep learning techniques are employed to train the
methods include KNN, support vector machine, nearest data in the healthcare sector, including convolutional encoder
subspace classification model, and representation networks, long short-term memories, CRF, U-Net CNN, and
classification model. Fuzzy C Means, hidden Markova WRN-PPNet [49]-[62].
random field, and self-organization map, are examples of D. PARAMETERS
unsupervised approaches [67]-[71]. MRI images are identified and categorized using an automated
C. DL MODELS technique, as described in the reference [31]. The Super Pixel
Deep learning (DL) models, as opposed to shallow Machine Methodology forms the foundation for this strategy, as does
Learning (ML) techniques, are founded on the principles of the categorization of every Super-the pixel. When attempting
learning data representations as well as learning hierarchical to categorize each superpixel as either tumorous or normal; the
features. Deep learning techniques are used to categorize brain extremely randomized trees classification model is evaluated
tumors, and these techniques find the descriptive data that alongside the support vector machine. This methodology
most properly describes the many forms of brain tumors. The utilizes two datasets, which are referred to as a dataset of
classification of brain tumors shifts away from being driven BRATS released in 2012 and 19 MRI FLAIR images
by manually created characteristics and toward being driven respectively. The findings indicate that the utilization of the
by data due to the nature of deep learning [87]. In the domain ERT classifier yields satisfactory results for this strategy. An
of deep learning technics, a convolutional neural network is instinctive organization technique is employed to recognize a
one of the most popularly utilized ones for the categorization tumor using a convolutional neural network with 3 x 3 tiny
of brain tumors, and a significant amount of progress has been kernels [84]. The tumor is ordered using this technique. The
made [88]. There are a few different approaches that may be method won the BRATS Challenge in 2013 by simultaneously
taken when classifying brain tumors, which can be seen in the placing first in the whole, core, and improving areas of the dice
research that was looked through. The difference includes the similarity, coefficient metric (0.880, 0.830, and 0.770). In the
following aspects: (i) the dataset that was used for [32], the Alexnet model convolutional neural network is
categorization, which included the types of tumors; (ii) the rummage-sale to concurrently identify multiple sclerosis and
pre-processing as well as data augmentation methods that normal tumors. A convolutional neural network was
were incorporated; (iii) The use of ROI segmentation as a successful in correctly categorizing 98.67% of the photos into
preliminary step in classification; The use of either a pre- one of three categories. To segment brain cancers from MRI
trained or custom-designed deep learning technique; and (iv) scans, a multi-stage Clustering framework was proposed in
the ROI segmentation question. For example, Bada and [54]. According to [85], there is a method for categorization
Barjaktarovi'c [88] used contrast-enhanced T1-weighted brain and segmentation that makes use of CNNs that is both efficient
tumor MRI images that were readily available to the public and effective. Image-Net was utilized in the suggested method
[89]. Meningioma, glioma, and pituitary tumor scans are in order to extract features. According to the findings, the
included in the collection, as well as images from the three classification was accurate to the extent of 97.5%, while the
anatomical perspectives of axial, sagittal, and coronal. The segmentation was accurate to the extent of 84%. In the study
images were preprocessed using several techniques, such as referred to as [86], Analysis of multiphase MRI images for
scaling and normalization, among others. To increase the size tumor grading has been conducted, and the results of base
of the training dataset, the photographs in the dataset are also neural networks and deep learning structures have been
flipped vertically and rotated via a 90-degree angle. contrasted and compared. According to the findings, the
Additionally, they employed a specially created CNN performance of the network which is measured by the
classifier trained with the Adam optimizer using a mini-batch specificity and sensitivity of CNN has increased by 18 percent
size of 16 and then tested the Classifier with 10-fold cross- when related to the efficiency of neural networks. In the paper
validation. A Glorot initializer is used to get the convolution [33], the authors present a deep learning-based supervised
layers' weights started off in the right direction. The measures technique for detection variations in artificial opening
utilized to assess the model's performance were the highly radar scans. This technique provides a dataset that had an
sensitive, selectivity, accuracy, recall, and F1-score. adequate amount of data volume and variety for the purpose
Meningiomas, gliomas, and pituitary tumors all have of training the DBN with the help of the input photographs and
sensitivity values of 89.8%, 96.2 percent, and 98.4%, the images that were acquired by applying structural
respectively. Meningiomas have a specificity of 90.2 percent, operatives on those images. The finding accuracy of this
gliomas have a specificity of 95.5 percent, and pituitary technique demonstrates the applicability of techniques based
tumors have a specificity of 97.7 percent according to the on deep learning for the purpose of finding solutions to change
model. In addition, the models have an overall accuracy of detection challenges.
95.4 percent, an average precision of 94.81 percent, an average
VOLUME XX, 2022 9
E. LIMITATIONS OF ML OVER DL APPROACHES fewest but most crucial features to further improve the
Recent research on the diagnosis of brain tumors is examined efficiency of the results. One of the most studied ideas in the
in this survey; the findings suggest that there is an opportunity field of deep learning is attention, which is used to solve issues
for further development in this area. Noise is introduced into like neural machine translation and image captioning. The
an MRI scan during the image capture process, and removing attention mechanism idea is supported by a number of
this noise is a complex process [20, 21–23]. Due to the theories, including Seq2Seq models, encoders, decoders,
tentacles and dispersed features that are characteristic of brain hidden states, context vectors, and others.
tumors [23]-[25], accurate segmentation is a thought- Some of the suggestions and possible improvements made by
provoking task. In order to achieve better categorization, one the published review articles includes: Further practice of
of the most significant tasks is to select and retrieve the hybrid-based learning technique is important to obtain strong
optimal features, as well as determine the right amount of CAD system [88], Noise estimation is challenging in machine
training and testing samples [26, 27]. The fact that deep learning and in deep learning lack of interpretability [89], How
learning models can autonomously learn new features is one effectively automatic methods can manage the impact of
of the reasons they are getting popular. On the other hand, treatment effects is still being researched [90], Technical
these models need a portion of memory and a lot of issues stemming from the difficulty in defining exactly what
dispensation control. Therefore, it is still necessary to develop deep learning is due to the lack of mathematical and theoretical
a lightweight computing framework that can produce a high foundations for many of its core models and techniques [91],
ACC in a shorter length of time. The following is a list of the Research should carefully consider how to lessen or
primary difficulties associated with detecting brain tumors. compensate for observer, spectrum, and selection biases, as
The glioma tumor and the stroke tumor do not contrast very well as how to increase reporting transparency [92], Research
well with one another. It is made up of tentacles and scattered should focus on optimization technique which will decide
components, both of which make the segmentation and number of layers and filters in the model [93], Semi supervised
categorization procedures far more difficult [83]. The training gives weak performance [94], Absences of transfer
identification of a tiny volume of the tumor remains difficult learning mechanism leads to weak generalization ability [95],
since it is possible for it to be recognized as a normal area [29]. Deficiency of training data and no resolution gives poor
Some of the existing approaches perform admirably for a performance of CNN [96], With large volume of data quality
complete tumor region but not for other regions (whether of image segmentation needed to improved [97], Transfer
enhanced or not), and conversely [90]-[95]. learning model is required incapacitating overfitting of image
F. FUTURE RESEARCH DIRECTIONS [98], Accurate analysis is difficult for vast number of images
This survey includes all of the significant features as well as [99], and Computation is difficult with multiple task [100].
the most recent work that has been done along with their
constraints and obstacles. The researchers will benefit from IV. PROPOSED FUTURE WORK
gaining a better understanding of how to do new research in The flowchart for the proposed work is given in Figure 3. This
the appropriate manner within a reasonable amount of time. describes the execution of the proposed system in the detection
Even though deep learning approaches have made substantial of various diseases using CNN. The entire architecture depicts
contributions, there is still a need for a generic approach. how the system deals with the recognition and detection of the
These methodologies produce better outcomes once training test image, and below we explain the process of execution.
and testing are carried out on achievement features (intensity The purpose of this research is to combine feature selection
range as well as resolution) that are comparable; moreover, the approaches with machine learning to identify pre-illnesses.
robustness of the methodologies are directly impacted by even For the early diagnosis of early diseases in MRI, CT scan, and
the slightest change between the training imaginings and the X-ray images, this system makes use of deep learning
testing imaginings. In the future, studies may be conducted to techniques and image processing technology. to make feature
detect brain cancers more precisely, with actual patient extraction more efficient, the dataset including defective
information since somewhat average (various image capture images from several categories was pre-processed and
methods) (scanners). Combining handcrafted characteristics segmented.
with deep features has the potential to enhance classification Image Acquisition: In image acquisition, heterogeneous
accuracy. Similar to this, lightweight technologies like images of the medical dataset collected which contains
quantum machine learning are crucial in enhancing accuracy abnormal and normal samples are gathered from a variety of
and efficacy, which in turn cuts down on the time required by individuals and converted into image format using a camera or
radiologists and raises the percentage of patients who survive some synthetic dataset.
their illnesses. An attention-based mechanism improves brain Pre-processing: There may be difficulties like noise, image
tumor segmentation outcomes and reduces computational blurring, and other concerns since the input data samples were
complexity issues. To be more precise, an image processing gathered from a range of people. As a consequence, pre-
and attention mechanism are used to extract the desired area processing methods are used for images in order to reduce
of the image, and then a pre-trained encoder part extracts the noise and improve image quality using modern techniques.
VOLUME XX, 2022 9
to a consistent size for effective preprocessing. The RGB

photos are first scaled and transformed to Hue Saturation
Intensity (HSI) format. Color perception is greatly aided by
the use of HSI color space representation. Masking is then
used to eliminate the pixels. Setting the pixel value of a picture
to zero or another background value is known as masking. The
diseased section of the original picture is then segmented using
the K-means segmentation technique. The goal of
segmentation is to transform a picture's representation into a
meaningful image that is simpler to explore. The best
characteristics from this dataset are then selected for accurate
categorization via feature selection. Relief-f Attribute
Evaluator (RAE), Principal Component Analysis (PCA), as
well as Information Gain (IG), are the three techniques of
feature selection used in this study.
Classification and Recognition: Disease classification is the
process of recognizing a test sample and giving it the
appropriate class label. The result of the feature extraction
module is to feed the classifier as an input. The classifier will
identify the right class label for the input image based on the
retrieved characteristics. There are a variety of methods for
categorization. Deep learning is one of them. Deep learning
employs a variety of artificial neural networks, including
CNN, ANN, RNN, and others. The image is sent into CNN,
which extracts the most important characteristics as distinct
layers. The key benefit of the convolutional neural network is
that it lowers the amount of work required by humans to
extract characteristics.
Implementation Process
1) The images from the dataset can be pre-processed in this
phase to prepare the data for subsequent processing. Initially,
photos are processed to reduce noise. The photos were then
transformed to grayscale and scaled to proper pixels while
keeping the aspect ratio constant.
2) The aspect ratio of the picture, the amount of lateral and
FIGURE 3. Flowchart of the proposed work
vertical lines in the picture, the location and number of loops
and curves, and other geometrical elements can all be
Processing the image is tough due to the fact that it is
extracted from each image sample through processing. These
originally in RGB color format. The RGB to greyscale
traits were then combined with the pixel-based information
conversion is required to reduce the complexity of a 3D pixel
from the image to yield accurate classification outcomes.
value to a 1D value. Many applications, such as edge
3) The input values of a traditional neural network are
detection, do not benefit from the use of three-dimensional
modified by passing through a sequence of hidden layers. A
pixels.
group of neurons makes up each layer, each of which is totally
Feature Selection: In image processing and data mining,
linked to every neuron in the layer preceding it. The superior
feature selection is critical. It calculates the best subset of
performance of CNNs is due to the fact that these networks
predicted characteristics from the original data. A subset of the
capture the fundamental features of pictures. This important
original characteristics is chosen that retains enough
property of CNN gives the confidence to apply it in the
information to distinguish successfully across classes. For
suggested dataset analysis.
feature selection, many search techniques can be utilized such
4) Download the dataset from open-source websites such as
as IG, PCA, and RAE.
the Kaggle dataset or any unseen dataset.
Feature Extraction: There are six separate sets of photos taken,
5) Extract various features by using the CNN model, train
from various available datasets. The obtained images are then
module, and save features in a. pkl file. There are below
subjected to image processing methods in order to identify
layered concepts in Convolutional Neural Networks:
valuable information for future study. Because the gathered
6) In the dense layer, it will classify the validation image set
photos are of various sizes, it is necessary to transform them
and show classification results accordingly.
VOLUME XX, 2022 9
V. CONCLUSION [10] Leonardo Rundo et al., “NeXt for neuro-radiosurgery: A fully

CAD systems for the detection of brain tumors are developed automatic approach for necrosis extraction in brain tumor MRI using an
unsupervised machine learning technique,” International Journal of
using brain MRI scans and digital image processing methods Imaging Systems and Technology, vol. 28, 2017, doi:
like pre-processing, separation, and classification. The classic https://doi:org/10.1002/ima.22253.
deep and machine-learning techniques for brain tumor [11] Jiang, Y. et.al., “A Brain Tumor Segmentation New Method Based on
identification are discussed in this work. Various research Statistical Thresholding and Multiscale CNN,” Intelligent Computing
publications from reputable journals and conferences have Methodologies, vol. 2, issue (3), pp. 235-245, 2019, doi:
https://doi:org/10.1007/978-3-319-95957-3_26.
been examined, with a full analysis of each work offered. This
section provides a summary of commonly used MRI datasets. [12] Liu, D. et.al., “3D Large Kernel Anisotropic Network for Brain Tumor
Segmentation,” Neural Information Processing, pp. 444-454, 2018, doi:
Although several ML and deep learning methods are used for https://doi:org/10.1007/978-3-030-04239-4_40.
classification, CNN has shown to be quite accurate. CNN is
[13] Nadeem, M. W. et.al., “Brain tumor analysis empowered with deep
often used to categorize brain tumors into two types: normal
learning: A review, taxonomy, and future challenges,” Brain Sciences, vol.
and pathological. The development of an autonomous brain 10, issue (2), pp. 118-151, 2020, doi: https://doi:org/
tumor detection system must consider reliability, accuracy, 10.3390/brainsci10020118.
and calculation time. This review examines current
methodologies and can be utilized in the future to build [14] Bhanothu, Y. et.al., “Detection and classification of brain tumor in mri
images using deep convolutional network,” International Conference on
effective diagnostic tools for additional brain illnesses that as Advanced Computing and Communication Systems , pp. 248–252, 2020,
Alzheimer's disease, Parkinson's disease, dementia, and stroke doi: https://doi:org/10.1109/ICACCS48705.2020.9074375.
using various MRI imaging modalities. Implementing this [15] Huang, Z. et.al., “Convolutional neural network based on complex
system in collaboration with multiple deep learning algorithms networks for brain tumor image classification with a modified activation
as deep hybrid learning for brain tumor detection and function,” IEEE Access vol. 8, pp. 89281–89290, 2020, doi:
https://doi:org/10.1109/ACCESS.2020.2993618.
classification will be future work for this study.
[16] Abiwinanda N. et.al., “Brain Tumor Classification Using
REFERENCES Convolutional Neural Network,” Medical Physics and Biomedical
[1] Machiraju Jaya Lakshmi & S.Nagaraja Rao., “Brain tumor magnetic Engineering, vol. 68, 2018, doi:https://doi:org/10.1007/978-981-10-9035-
resonance image classification: a deep learning approach,” Soft Computing, 6_33.
vol. 26, pp.6245–6253, 2022, doi:https://doi:org/10.1007/s00500-022- [17] Ari, A.; Hanbay, D., “Deep learning-based brain tumor classification
07163-z. and detection system,” Turk. J. Electr. Eng. Comput. Sci, vol. 26, pp. 2275–
2286, 2018, doi: http://doi:org/10.3906/elk-1801-8.
[2] Wen Jun & Zheng Liyuan, “Brain Tumor Classification Based on
Attention Guided Deep Learning Model,” International Journal of [18] Ishikawa, Y. et.al., “Brain tumor classification of microscopy images
Computational Intelligence Systems, vol.15, pp. 35, 2022, doi: using deep residual learning,” SPIE BioPhotonics Australas. 2016, vol.
https://doi:org/10.1007/s44196-022-00090-9. 10013, 100132Y, doi: https://doi:org/10.1117/12.2242711.
[3] Arshia Rehman et.al., “A Deep Learning-Based Framework for [19] Mohsen, H. et.al., “Classification using deep learning neural networks
Automatic Brain Tumors Classification Using Transfer Learning.” Circuits for brain tumors,” Futur. Comput. Inform. J. vol. 3, issue 1, pp. 68–71, 2018,
Systems and Signal Processing, vol. 39, pp. 757–775, 2020, doi: doi: https://doi:org/10.1016/j.fcij.2017.12.001.
https://doi:org/10.1007/s00034-019-01246-3. [20] Paul, J. et.al., “Deep learning for brain tumor classification,”
Biomedical Applications in Molecular, vol. 10137, pp. 1013710-1013726,
[4] Fernando, T. et.al., “Deep Learning for Medical Anomaly Detection – A 2017, doi: https://doi:org/10.1117/12.2254195.
Survey,” ACM Computing Surveys, vol. 54, issue (7), pp. 1-37, 2021, doi:
https://doi:org/10.1145/3464423. [21] Xu, Y. et.al., “Deep convolutional activation features for large scale
brain tumor histopathology image classification and segmentation’” IEEE
[5] Alexander Selvikvåg Lundervold & Arvid Lundervold, “An overview of International Conference on Acoustics, Speech and Signal Processing
deep learning in medical imaging focusing on MRI,” Zeitschrift für (ICASSP), Brisbane, Australia, 19–24 April 2015; pp. 947–951, 2015, doi:
Medizinische Physik, vol. 29, issue (2), 2019, pp. 102-127, doi: https//doi:org/ 978-1-4673-6997.
https://doi:10.1016/j.zemedi.2018.11.002.
[22] Ahmed et.al., “Fine-tuning convolutional deep features for MRI based
brain tumor classification,” Medical Imaging, Vol. 10134, 2017, doi:
[6] Rundo, L. et al., “Semi-automatic Brain Lesion Segmentation in Gamma
https://doi:org/10.1117/12.2253982.
Knife Treatments Using an Unsupervised Fuzzy C-Means Clustering
Technique,” Smart Innovation, Systems and Technologies, vol 54. Springer, [23 Ismael, Mustafa Rashid. Et.al., "Hybrid Model - Statistical Features and
Cham. 2016, doi: https://doi:org/10.1007/978-3-319-33747-0_2. Deep Neural Network for Brain Tumor Classification in MRI Images”
Dissertations. Vol. 3291, 2018 https://scholarworks.wmich.
[7] Stijn Bonte et al., “Machine learning based brain tumour segmentation
edu/dissertations/3291.
on limited data using local texture and abnormality,” Comput Biol Med, vol.
98, pp. 39-47, 2018, doi: https://doi:10.1016/j.compbiomed.2018.05.005. [24] Liu, R.et.al., “ Exploring deep features from brain tumor magnetic
resonance images via transfer learning,” International Joint Conference on
[8] Carmelo Militello et al., “Gamma Knife treatment planning: MR brain
Neural Networks (IJCNN), pp. 235–242, 2016, doi: https://doi:org/
tumor segmentation and volume measurement based on unsupervised Fuzzy
10.1109/IJCNN.2016.7727204.
C-Means clustering,” International Journal of Imaging Systems and
Technology, vol. 25 (3), pp. 213-225, 2015, [25] Ladefoged, C.N. et.al., “Deep learning-based attenuation correction of
doi: https://doi:org/10.1002/ima.2213. PET/MRI in pediatric brain tumor patients: Evaluation in a clinical setting,”
Front. Neurosci, vol. 2, 2018, doi: https://doi:org/10.3389/fnins.
[9] J Juan-Albarracín, et al., “Automated Glioblastoma Segmentation Based
2018.01005.
on a Multiparametric Structured Unsupervised Classification,” PLoS ONE,
vol. 10, issue (5), 2015, doi: https://doi:org/10.1371/journal.pone.0125143.
VOLUME XX, 2022 9
[26] Fabelo, H.et.al., “Deep learning-based framework for in vivo [43] Ramin Ranjbarzadeh.et.al. “Brain tumor segmentation based on deep
identification of glioblastoma tumor using hyperspectral images of human learning and an attention mechanism using MRI multi‑modalities brain
brain,” Sensors, vol. 19, issue 4, Doi: https://doi:org/10.3390/s19040920. images,” Scientific Reports, 2021, vol. 11, pp. 10930, 2021, doi:
https://doi:org/10.1038/s41598-021-90428-8.
[27] Li, Y & Shen, L., “Deep learning based multimodal brain tumor
diagnosis,” In International MICCAI Brainlesion , vol. 10670, pp. 149–158, [44] T. Ruba1.et.al.” Brain tumor segmentation in multimodal MRI images
2017, doi: https://doi:org/ https://doi:org/10.1007/978-3-319-75238-9_13. using novel LSIS operator and deep learning,” Journal of Ambient
[28] Chato, L.& Latifi, S., “Machine learning and deep learning techniques Intelligence and Humanized Computing, 2022, doi:
to predict overall survival of brain tumor patients using MRI images,” https://doi:org/10.1007/s12652-022-03773-5.
International Conference on Bioinformatics and Bioengineering, vol. 2017
pp. 9–14, 2017, doi: http://dx.doi:org/10.1109/BIBE.2017.00-86. [45] Aman Verma1 &Vibhav Prakash Singh, “Design, analysis and
implementation of efficient deep learning frameworks for brain tumor
[29] Amarapur, B., “Computer-aided diagnosis applied to MRI images of classification,” Multimedia Tools and Applications, vol. 81, pp. 37541–
brain tumor using cognition based modified level set and optimized ANN 37567, 2022, doi: https://doi:org/10.1007/s11042-022-13545-0.
classifier,” Multimed. Tools Appl, vol. 3601, issue 2020, doi:
https://doi:org/10.1007/s11042-018-6308-7. [46] Praveen Kumar Ramtekkar et.al. “Innovative brain tumor detection
using optimized deep learningTechniques,” Int J Syst Assur EngManag,
[30] Benson, E. et.al., “Deep Hourglass for Brain Tumor Segmentation,” 2022 doi: https://doi:org/10.1007/s13198-022-01819-7.
International MICCAI Brainlesion Workshop, Vol. 10, issue (2), pp. 419–
428,2018 doi: https:// https://doi:org/10.3390/brainsci10020118. [47] R. Rajasree1 et.al.” Multiscale-based multimodal image classification
[31] Zhou, C. et.al “Learning contextual and attentive information for brain of brain tumor using deep learning method,” Neural Computing and
tumor segmentation,” International MICCAI Brainlesion Workshop, vol. Applications, vol. 33, pp.5543–5553, 2021, doi:
4798, pp. 497–507,2018 doi: https://doi:org/10.1007/978-3-030-11726- https://doi:org/10.1007/s00521-020-05332-5.
9_44.
[48] Tuhin MA.et.al., “Detection and 3d visualization of brain tumor using
[32] Kim G.et.al.,”Brain Tumor Segmentation Using Deep Fully deep learning and polynomial interpolation,” IEEE Asia-Pacific Conference
Convolutional Neural Networks,” Brainlesion, vol. 10670,2018, doi: on Computer Science and Data Engineering, pp. 1-6, 2020, doi:
https://doi:org/10.1007/978-3-319-75238-9_30. https://doi:org/10.1109/ChSDE50874.2020.9411595.
[33] Afshar, P. et.al., “Brain tumor type classification via capsule [49] Shirwaikar RD. et.al.,” A survey on Brain Tumor Detection using
networks,” IEEE International Conference on Image Processing, pp. 3129– Machine Learning,” International Conference on Forensics, vol. 1, pp. 1-6,
3133, 2018, doi: https://doi:org/10.1109/ICIP.2018.8451379. 2021 doi : https://doi:org/10.1109/FABS52071.2021.9702583.
[34] Chang, P.D., “Fully convolutional deep residual neural networks for
brain tumor segmentation,” International Workshop on Brainlesion, pp. [50] H. Yahyaoui.et.al., ‘‘Deep learning guided by an ontology for medical
108–118, 2016, doi: https://doi:org/10.1007/978-3-319-55524-9_11. images classification using a multimodal fusion,’’ Adv. Technol. Eng, pp.
1–6, 2021, doi: https://doi:org/10.1109/ICOTEN52080.2021.9493469.
[35] Isensee, F. et.al., “Brain tumor segmentation using large receptive field
deep convolutional neural networks,” Medizin , pp. 86–91, 2017, doi: [51] S. Pokhrel. et.al., "Brain Tumor Detection Application Based on
https://doi:org/10.1007/978-3-662-54345-0_24. Convolutional Neural Network," International Conference on Intelligent
Technologies (CONIT), pp.1-5, 2021, doi: https://doi: 10.1109/
[36] Kumar, S.et.al.,“Semantic Segmentation Using Deep Learning for
CONIT55038.2022.9848177.
Brain Tumor MRI via Fully Convolution Neural Networks,” Information
and Communication Technology for Intelligent Systems; Springer, pp. 11–
[52] S. Gull.et.al., "An Interactive Deep Learning Approach for Brain
19, 2019, doi: https://doi:org/10.1007/978-981-13-1742-2_2.
Tumor Detection Through 3D-Magnetic Resonance Images," International
[37] AM. Hasan.et.al., “Combining Deep and Handcrafted Image Features Conference on Frontiers of Information Technology, 2021, pp. 114-119,
for MRI Brain Scan Classification,” IEEE Access, vol. 7, pp. 79959–79967, 2021, doi: https://doi:10.1109/FIT53504.2021.00030.
2019, doi: https://doi:org/10.1109/ACCESS.2019.2922691.
[53] Suter Y et.al.,” Deep Learning versus Classical Regression for Brain
[38] S. Deepak & PM. Ameer, “Brain tumor classification using deep CNN
Tumor Patient Survival Prediction,” International MICCAI Brainlesion
features via transfer learning,” Computers in Biology and Medicine, pp.1-
Workshop, pp. 429–440,2019, doi: https://doi:10.1007/978-3-030-11726-
7,2019, doi: https://doi:org/10.1016/j.compbiomed.2019.103345.
9_38.
[39] Rai, Hari &Chatterjee, Kalyan. “2D MRI image analysis and brain [54] Hu Y&Xia, Y.,”3D deep neural network-based brain tumor
tumor detection using deep learning CNN model LeU-Net.”, Multimedia segmentation using multimodality magnetic resonance sequences.”
Tools and Applications, Vol. 80, pp. 36111-36141, 2021, doi: http://doi:org/ International MICCAI Brainlesion Workshop, pp. 423–434, 2017, doi:
10.1007/s11042-021-11504-9. https://doi:org/10.1007/978-3-319-75238-9_36.
[40] Deepak V.K & Sarath R. “An intelligent brain tumor segmentation [55] Nie, D. et.al.,” 3D deep learning for multi-modal imaging-guided
using improved Deep Learning Model Based on Cascade Regression survival time prediction of brain tumor patients,” International Conference
method,” Multimedia Tools and applications, 2022, doi: https://doi:org on Medical Image Computing and Computer-Assisted Intervention, pp.
/10.1007/s11042-022-13945-2. 212–220, 2016, doi: https://doi:10.1007/978-3-319-46723-8_25.
[56] G. Anand kumar & P.shridevi.,” 3D Deep Learning for Automatic
[41] Haq, E.U. et al. “An MRI-based deep learning approach for efficient
Brain MR Tumor Segmentation with T-Spline Intensity Inhomogeneity
classification of brain tumors,” J Ambient Intell Human Comput , 2021, doi:
Correction,” Automatic Control and Computer Sciences ,Vol. 52, pp. 439–
450, 2018, doi:https://doi:org/10.3103/S0146411618050048.
[42] Sidra Sajid.et.al. “BrainTumor Detection and Segmentation in MR
[57] Girija Chetty. et.al., “A low resource 3D U-Net based deep learning
Images Using Deep Learning,” Arabian Journal for Science and
model for medical image analysis.” Int. j. inf.tecnol.vol.14, pp. 95–103,
Engineering, vol. 44, pp. 9249–9261, 2019, doi:
2022, doi: https://doi:org/10.1007/s41870-021-00850-4.
[58] Zeeshan Shaukat et. al. “A state‑of‑the‑art technique to perform
cloud‑based semantic segmentation using deep learning 3D U‑Net
VOLUME XX, 2022 9
architecture,” BMC Bioinformatics, Vol. 23, pp. 251-272, 2022, doi: [74] Banerjee, S. et.al.,” Deep Radiomics for Brain Tumor Detection and
https://doi:org/10.1186/s12859-022-04794-9. Classification from Multi-Sequence MRI,” SN Computer Science, vol. 1,
issue, 4, 2019, doi: https://doi:org/10.1007/s42979-020-00214-y.
[59] Hiba Mzoughi et.al. “Deep Multi-Scale 3D Convolutional Neural
Network (CNN) for MRI Gliomas Brain Tumor Classification,” Journal of [75] Naseer, A. et.al.,” Computer-Aided Brain Tumor Diagnosis:
Digital Imaging, vol.33, pp. 903–915, Performance Evaluation of Deep Learner CNN Using Augmented Brain
2020,doi:https://doi:org/10.1007/s10278-020-00347-9. MRI,” International Journal Of Biomedical Imaging, vol. 2021, pp. 1-11,
2021, doi: https://doi:org/10.1155/2021/5513500.
[60] Joseph Stember& Hrithwik Shalu.,” Deep Reinforcement Learning [76] Abd El Kader et.al.,” Brain Tumor Detection and Classification on MR
with Automated Label Extraction from Clinical Reports Accurately Images by a Deep Wavelet Auto-Encoder Model,” Diagnostics, vol. 11,
Classifies 3D MRI Brain Volumes,” Journal of Digital Imaging, vol. 35, 2021, doi:https://doi:org/10.3390/diagnostics11091589.
pp. 1143–1152,2022, doi: https://doi:org/10.1007/s10278-022-00644-5.
[77] Sheeba, L. et.al.,” Detection of Exact Location of Brain Tumor from
[61] Yohan Jun.et.al., “Deep-learned 3D black-blood imaging using MRI Data Using Big Data Analytics,” Systematic Review Pharmacy, vol.
automatic labelling technique and 3D convolutional neural networks for 12(5), pp. 378-381, 2021. doi:http://doi: 10.5530/srp.2020.1.01.
detecting metastatic brain tumors,” Scientific Report, Vol. 8, pp. 9450-9461,
2018, doi:https:// doi:10.1038/s41598-018-27742-1. [78] Amin J,et.al.,”Brain tumor detection and classification using machine
learning: a comprehensive survey,” Complex & Intelligent Systems, vol. 8,
[62] Pranjal Agrawal.et.al.,” Segmentation and classification of brain tumor pp. 3163-3183, 2022, doi:https://doi:org/10.1007/s40747-021-00563-y.
using 3D-UNet deep neural networks,” International Journal of Cognitive [79] Shahajad, M.et.al.,”Features extraction for classification of brain tumor
Computing in Engineering, vol. 3, pp. 199–210, 2022, doi: mri images using support vector machine,” Cloud Computing, Data Science
https://doi:org/10.1016/j.ijcce.2022.11.001. Engineering (Confluence), pp. 767– 772, 2021
doi:https://doi:10.1109/Confluence51648.2021.9377111.
[63] Agus Subhan Akbar.et.al., “Single level UNet3D with multipath
residual attention block for brain tumor segmentation,” Computer and [80] Alberts, E.et.al.,”Multi-modal image classification using low-
Information Sciences, vol. 34, pp. 3247–3258, dimensional texture features for genomic brain tumor recognition,”
2022,doi:https://doi:org/10.1016/j.jksuci.2022.03.022. Biomedical Image Analysis, vol. 10551, pp. 201–209, 2017, doi:
https://doi: 10.1007/978-3-319-67675-3_18.
[64] HT. Zaw.et.al., “Brain tumor detection based on Naïve Bayes
classification,” Applied Sciences andTechnology, pp. 1-4, 2019, doi: [81] Iqbal, S.et.al.,”Computer-assisted brain tumor type discrimination
https://doi:10.1109/ICEAST.2019.8802562. using magnetic resonance imaging features,” Biomed. Eng, Vol. 8, pp. 5–
28, 2018, doi: https://doi:org/10.1007/s13534-017-0050-3.
[65] E. Sert et.al., “A new approach for brain tumor diagnosis system: Single [82] Deepa, A.R.& Emmanuel, W.R.S.,” An efficient detection of brain
image super resolution based maximum fuzzy entropy segmentation and tumor using fused feature adaptive firefly backpropagation neural network,”
convolutional neural network," Medical Hypothesis, vol. 133, pp. 1-9, 2019, Multimed. Tools Appl, vol. 78, issue (11), pp. 11799–11814, doi:
doi :https://doi:org/10.1016/j.mehy.2019.109413. https://doi:org/10.1007/s11042-018-6731-9.
[66] TL. Narayana & TS.Reddy., “An Efficient Optimization Technique to [83] Amin, J. et.al.,” Detection of Brain Tumor based on Features Fusion
Detect Brain Tumor from MRI Images,” International Conference on Smart and Machine Learning,” Journal of Ambient Intelligence and Humanized
Systems and Inventive Technology, pp. 1-4, 2018. doi: https:doi: Computing, 2018, https://doi:doi:org/10.1007/s12652-018-1092-9.
10.1109/ICSSIT.2018.8748288.
[84] McKinley.et.al.,” Fully convolutional networks with dense skip
connections for semantic segmentation, with application to brain tumor
[67] FP. Polly et.al.,” Detection and Classification of HGG and LGG Brain segmentation,” MICCAI Brainlesion Workshop, vol. 10670, pp. 169–177,
Tumor Using Machine Learning,” International conference on Information 2018, doi: https://doi: 10.1007/978-3-319-75238-9_15.
Networking, pp. 813-817, 2018, doi:https://10.1109/ICOIN.2018.834323.
[85] Natarajan, A& Kumarasamy, S.” Efficient Segmentation of Brain
[68] J.Amin et.al., “A distinctive approach in brain tumor detection and Tumor Using FL-SNM with a Metaheuristic Approach to Optimization,”
classification using MRI,” Pattern Recognition Letters, vol. 139, pp. 1-10, Journal of Medical Systems, vol. 43, pp. 25, 2019, doi:
2017, doi:https://doi:10.1016/j.patrec.2017.10.036. https://doi.org/10.1007/s10916-018-1135-y.
[69] N. Gupta &P. Khanna,.“A non-invasive and adaptive CAD system to [86] Ben naceur et.al.,” Fully automatic brain tumor segmentation with deep
detect brain tumor from T2-weighted MRIs using customized Otsu’s learning-based selective attention using overlapping patches and multi-class
thresholding with prominent features and supervised learning,” Signal weighted cross-entropy,” Medical Image Analysis, vol. 63, 2020, doi:
Processing: Image Communication, vol. 59, pp. 18-26, 2017, https://doi:org/10.1016/j.media.2020.101692.
doi:https://doi:org/10.1016/j.image.2017.05.013.
[87] Amiri, S. et.al.”Bayesian Network and Structured Random Forest
[70] N. Gupta et.al.,” Identification of Gliomas from brain MRI through Cooperative Deep Learning for Automatic Multi-label Brain Tumor
adaptive segmentation and run length of centralized patterns,” Journal of Segmentation,” ICAART, vol.2, pp. 183–190, 2018, doi:https:// doi:
Computational Science, vol. 25, pp. 1-8, 2017, doi:https://doi: 10.5220/0006629901830190
10.1016/j.jocs.2017.02.009.
[88] Saqib Ali.et.al.” A Comprehensive Survey on Brain Tumor Diagnosis
[71] A.Minz & C. Mahobiya ,”MR Image classification using adaboost for Using Deep Learning and Emerging Hybrid Techniques with Multi‑modal
brain tumor type,” International Advance Computing Conference, pp. 1-5, MR Image,” Archives of Computational Methods in Engineering , vol. 29,
2017, doi:https://doi: 10.1109/IACC.2017.0146. pp. 4871–4896, 2022, doi:https://doi:org/10.1007/s11831-022-09758-z.
[72] AS. Shankar et.al.,”Brain Tumor Classification Using Gustafson–
kessel (G-k) Fuzzy Clustering Algorithm,” International Journal of Latest [89] Erena Siyoum Biratu. et.al., “A Survey of Brain Tumor Segmentation
and Classification Algorithms”, J. Imaging, vol. 7, 2021, doi:
Engineering Research and Applications, vol. 1, pp. 68-72, 2016.
https://doi:org/10.3390/jimaging7090179.
[73] Islam, M.et.al.,”Brain tumor detection in MR image using superpixels,
principal component analysis and template based K-means clustering [90] Stefan Bauer et. al. “A survey of MRI-based medical image analysis
algorithm,” Machine Learning With Applications, vol. 5, 2021, for brain tumor studies,” Physics in medicine and biology, vol. 58, pp. 97–
doi:https://doi:org/10.1016/j.mlwa.2021.100044. 129, 2013, doi: https:// doi:10.1088/0031-9155/58/13/R97.
VOLUME XX, 2022 9
[91] C.S. Rao & K. Karunakara, “A comprehensive review on brain tumor UDAY PRATAP SINGH obtained B.Sc. Degree
segmentation and classification of MRI images,” Multimed Tools Appl, vol. from Dr. R.M.L. Awadh University, Faizabad,
80, pp. 17611–17643, 2021, doi:https:// 10.1007/s11042-020-10443-1. (U.P.), India, M.Sc. Degree from Indian Institute
of Technology, Guwahati, India, and received
[92] Ramya Balakrishnan.et.al “Automatic segmentation of white matter Ph.D. in Computer Science from Barkatullah
hyperintensities from brain magnetic resonance images in the era of deep University, Bhopal. He is currently working as an
learning and big data – A systematic review,” Computerized Medical Associate Professor in the Department of
Imaging and Graphics, vol.88, pp. 101867-101888,2021, Mathematics Shri Mata Vaishno Devi University
doi:https://doi:org/10.1016/j.compmedimag.2021.101867. Katra, 182320, Jammu and Kashmir, India. He had
worked on various areas such as Soft Computing,
[93] P Gokila Brindha.et.al.” Brain tumor detection from MRI images using Theoretical Computer Science, and Image
deep learning Ttechniques,” Materials Science and Engineering, Processing. He had authored several research papers, published at reputed
vol.1055,pp.1-8, 2021,doi:https://doi:10.1088/1757-899X/1055/1/012115. journals and conferences. He is a life member of IAENG, and Computer
Society of India.
[94] Zhihua Liu.et.al.” Deep learning-based brain tumor segmentation: a
survey,” Complex and Intelligent Systems, 2022, doi: SIDDHARTH SINGH CHOUHAN have
https://doi:org/10.1007/s40747-022-00815-5. completed Ph.D. in Computer Science and
Engineering from Shri Mata Vaishno Devi
[95] Zeynettin Akkus.et.al.” Deep Learning for Brain MRI Segmentation: University Katra, 182320, Jammu and Kashmir,
State of the Art and Future Directions,” J Digit Imaging, vol. 30, pp. 449– India. He is a PostDoc from the University of
459, 2017,doi: https://doi:10.1007/s10278-017-9983-4. Malta, Europe. He has received B.E. degree in
2010 and M.Tech degree in 2013 in Computer
[96] Parvathy Jyothi. & A. Robert Singh. “Deep learning models and Science and Engineering from RGPV University,
traditional automated techniques for brain tumor segmentation in MRI: a Bhopal, India. His area of interest includes
review,” Artificial Intelligence Review, 2022, doi: developing Soft Computing, Computer vision,
https://doi:org/10.1007/s10462-022-10245-x. Drone Technology, Deep Learning, and Precision
Agriculture. He had authored several research papers, published at reputed
[97] Sindhu Devunooru.et.al.” Deep learning neural networks for medical journals and conferences.
image segmentation of brain tumours for diagnosis: a recent review and
taxonomy,” Journal of Ambient Intelligence and Humanized Computing, SANJEEV JAIN born in Vidisha, Madhya
vol. 12, pp. 455–483, 2021, doi:https://doi.org/10.1007/s12652-020-01998- Pradesh, in 1967. He received the master’s degree
w. in Computer Science and Engineering from IIT
Delhi, New Delhi, India, in 1992, and the Ph.D.
[98] Huiyan Jiang.et.al.” Deep learning techniques for tumor segmentation: degree in Computer Science and Engineering. He
a review,” The Journal of Supercomputing ,vol.78, pp.1807–1851, 2022, has over 33 years’ experience in teaching and
doi:https://doi.org/10.1007/s11227-021-03901-6. research. He is currently a Vice Chancellor of
Central University of Jammu. He has the credit of
[99] Sabaa Ahmed Yahya Al‑Gala.et.al.” MRI brain tumor medical images making significant contribution to research and
analysis using deep learning techniques: a systematic review,” Health and development in the area of image processing and
Technology, vol. 11, pp. 267–282, 2021, doi: mobile adhoc network. He has guided Ph.D.
https://doi:org/10.1007/s12553-020-00514-6. Scholars and has undertaken a number of major Research and Development
projects sponsored by the Government and Private Agencies. He is a
[100] Arkapravo Chattopadhyay & Mausumi Maitra.” MRI-based brain member of various societies like Computer Society of India, IEEE.
tumour image detection using CNN based deep learning method,”
Neuroscience Informatics, vol.2, pp.100060-100066, 2022, doi:
https://doi:org/10.1016/j.neuri.2022.100060.
SHUBHANGI SOLANKI received B.E

Computers in 2002 from Savitribai Phule Pune
University (SPPU), received M.Tech in Artificial
intelligence from RGPV, Bhopal in 2007 with Hons. She has published
more than 10 papers and currently working as Associate professor in
Computer Engineering Department with SIT, Lonavala with more than
10yrs.
VOLUME XX, 2022 9

Brain Tumor Detection and Classification Using Int

Uploaded by

Brain Tumor Detection and Classification Using Int

Uploaded by

This article has been accepted for publication in IEEE Access.

Brain Tumor Detection and Classification using

I. INTRODUCTION because surgery alone would be insufficient to treat them.

VOLUME XX, 2017 1

Reference Methodology Algorithms Accuracy Dataset used

2 VOLUME XX, 2017

2 VOLUME XX, 2017

weighted, FLAIR, and ADM from the H.

2 VOLUME XX, 2017

Reference Methodology Algorithms Accuracy Dataset used

2 VOLUME XX, 2017

B. MACHINE LEARNING METHODS

Reference Methodology Algorithms Accuracy Dataset used

2 VOLUME XX, 2017

Reference Methodology Algorithms Gap Analysis Dataset used

2 VOLUME XX, 2017

2 VOLUME XX, 2017

III. FINDINGS global thresholding. Deep learning-based segmentation allows

2 VOLUME XX, 2017

are used to classify data. multi-kernel SVM Binary

study. The deep learning approach can extract more detailed 90

and normalization to enhance desired features. Pre-processing

VOLUME XX, 2022 9

VOLUME XX, 2022 9

VOLUME XX, 2022 9

to a consistent size for effective preprocessing. The RGB

VOLUME XX, 2022 9

V. CONCLUSION [10] Leonardo Rundo et al., “NeXt for neuro-radiosurgery: A fully

VOLUME XX, 2022 9

VOLUME XX, 2022 9

VOLUME XX, 2022 9

SHUBHANGI SOLANKI received B.E

VOLUME XX, 2022 9

You might also like