Hindawi

Journal of Sensors
Volume 2023, Article ID 1224619, 19 pages
https://doi.org/10.1155/2023/1224619

Research Article
An Effective and Novel Approach for Brain Tumor Classification
Using AlexNet CNN Feature Extractor and Multiple Eminent
Machine Learning Classifiers in MRIs

Alok Sarkar ,1 Md. Maniruzzaman ,1 Mohammad Ashik Alahe ,1

and Mohiuddin Ahmad 2
1
Electronics and Communication Engineering Discipline, Khulna University, Khulna, 9208, Bangladesh
2
Department of Electrical and Electronic Engineering, Khulna University of Engineering & Technology, Khulna, 9203, Bangladesh

Correspondence should be addressed to Alok Sarkar; alokece99@gmail.com

Received 27 October 2022; Revised 28 January 2023; Accepted 27 February 2023; Published 8 March 2023

Academic Editor: Akhilesh Pathak

Copyright © 2023 Alok Sarkar et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

A brain tumor is an uncontrolled malignant cell growth in the brain, which is denoted as one of the deadliest types of cancer in
people of all ages. Early detection of brain tumors is needed to get proper and accurate treatment. Recently, deep learning
technology has attained much attraction to the physicians for the diagnosis and treatment of brain tumors. This research
presents a novel and effective brain tumor classification approach from MRIs utilizing AlexNet CNN for separating the dataset
into training and test data along with extracting the features. The extracted features are then fed to BayesNet, sequential
minimal optimization (SMO), Naïve Bayes (NB), and random forest (RF) classifiers for classifying brain tumors as no-tumor,
glioma, meningioma, and pituitary tumors. To evaluate our model’s performance, we have utilized a publicly available Kaggle
dataset. This paper demonstrates ROC, PRC, and cost curves for realizing classification performance of the models; also,
performance evaluating parameters, such as accuracy, sensitivity, specificity, false positive rate, false negative rate, precision, f-
measure, kappa statistics, MCC, ROC area, and PRC area, have been calculated for four testing options: the test data itself,
cross-validation fold (CVF) 4, CVF 10, and percentage split (PS) 34% of the test data. We have achieved 88.75%, 98.15%,
86.25% and 100% of accuracy using the AlexNet CNN+BayesNet, AlexNet CNN+SMO, AlexNet CNN+NB, and AlexNet CNN
+RF models, respectively, for the test data itself. The results imply that our approach is outstanding and very effective.

1. Introduction structure of the brain to diagnose tumors. The most popular

The existence of any kind of abnormalities in the brain
may lead the human nervous system as well as health in
great danger [1]. Today, brain tumor is denoted as the
most severe ones, which results from the development of
uncontrolled destructive cells in the human brain. Accord-
ing to the World Health Organization and the American
Brain Tumor Association reports [2], it can be categorized
as glioma, meningioma and pituitary tumors based on cell
structure and locations.
It is necessary to completely identify the types of brain
tumors for proper diagnosis and treatment. The radiologist
utilizes various clinical imaging techniques to visualize the
imaging techniques are ultrasound imaging, X-ray imaging,
positron emission tomography, computed tomography,
and magnetic resonance imaging (MRI). Among these, the
MRI provides precise information on a given medical image,
gives definite information on the shape of the cerebrum
structure, and the location of inconsistencies in brain tissues.
However, it is often chaotic, time-consuming, and error-
prone if medical specialists manually detect brain tumors
by examining MRIs [3]. To diminish these problems, the
computer-aided detection and diagnostic approach has been
developed for the last few decades [4–11]. In recent years,
various efforts have been taken to elevate a robust and
precise technique to classify brain tumors automatically by
2 Journal of Sensors
using machine learning in MRIs [12–14]. Nevertheless,
because of high inter- and intrashape, texture, and variation
of contrast, this still remains a challenging task. Conven-
tional machine learning techniques rely on artisanal features,
which limit the robustness of the outcomes, while deep
learning techniques automatically extract important func-
tionality that provides significantly better performance.
But still, the results obtained using deep learning tech-
nologies are not sufficient, and they do not always show
good classification accuracy, uses of only AlexNet, for exam-
ple [15]. The research is challenging, especially for selecting
the region of interest (ROI), extracting deep features, using
clamorous images, etc.
To minimize the aforementioned problems, in this
study, we have proposed a novel and effective technique
for brain tumor classification automatically employing the
AlexNet CNN feature extractor and a variety of prominent
machine learning classifiers, such as BayesNet, sequential
minimal optimization (SMO), Naïve Bayes (NB), and ran-
dom forest (RF) in MRIs.
We know that AlexNet CNN is a good tool to determine
ROI for extracting deep features and can easily deal with
clamorous and raw images, which are very essential for the
current research. Moreover, deploying machine learning
classification algorithms with AlexNet CNN certainly
enhance the performance of the proposed model.
Our contributions to this work are as follows:
(i) We have utilized AlexNet CNN for separating
human brain MRI datasets into training and test
data and also for extracting the features from those
separated datasets
(ii) Those extracted features are then used for classifying
the brain tumors as no-tumor, glioma, meningioma,
and pituitary tumors by using eminent BayesNet,
SMO, NB, and RF classifiers
This paper is organized as follows. Section 1 discusses
briefly the problems, some of the existing solutions, research
gaps, objectives, contributions, and organizations of the paper.
Section 2 presents related works on brain tumor detection and
classification. It also discusses the research gaps at the end of
this section. Section 3 describes overall working methodology,
proposed architecture, and performance measuring metrics.
Section 4 describes the results and performance of our model
along with a comparison with other contemporary findings.
Finally, Section 5 contains the conclusion of the entire study,
limitations, and future works.
2. Related Works
Many researchers have published their works to represent
different methodologies or approaches for detection and
classification purposes over the past few years [15–36].
Various methodologies have been tested using different clin-
ical databases, including magnetic resonance imaging (MRI)
of brain tumors. Selvaraj et al. [37] developed a binary
classifier for classifying brain MRIs a utilizing first- and
second-order statistics and a least square support vector
machine (SVM) classifier. Sarkar et al. [38] proposed a
computer-aided approach for detecting and classifying brain
tumors from MRIs utilizing genetic algorithm as a feature
extractor and SVM as classifier, where they obtained 98.3%
classification accuracy. John [25] used techniques based on
the gray level cooccurrence matrix (GLCM) and discrete
wavelet transform (DWT) to detect and classify brain
tumors. Ullah et al. [39] applied DWT for feature extraction
and a feed-forward artificial neural network for brain MRI
classification. Kharrat et al. [40] classified brain tumors as nor-
mal and abnormal classes through genetic algorithms and
SVM classifiers. Papageorgiou et al. [33] obtained 90.26%
and 93.22% accuracy for low- and high-grade glioma tumors,
respectively, through fuzzy cognitive maps. Díaz-Pernas et al.
[26] proposed an architecture for automatic detection and seg-
mentation of brain tumors and acquired 97.3% accuracy.
Shree and Kumar [41] used a GLCM feature extractor and
probabilistic neural network classifier for classifying MRI
datasets and achieved 95% accuracy. Arunachalam and Savar-
imuthu [42] offered an architecture that covered enhancement
of image, transformation of image, extraction of features, and
finally, image classification. They employed shift-invariant
shear-let transforms for enhancing MRIs and used numerous
feature extractors, such as Gabor, GLCM, and DWT to extract
the features. Hossain et al. [30] utilized the traditional CNN
approach to classify brain tumor images. They utilized fuzzy
c-means clustering for segmenting the tumors and achieved
an accuracy of 97.87%.
Several deep learning approaches have been extensively
utilized by researchers to detect and classify brain tumors
in the last decades. Sajid et al. [43] proposed a fully auto-
mated hybrid approach for the detection and segmentation
of brain tumors in MRIs using deep learning. They utilized
the BRATS2013 dataset and obtained 86%, 86%, and 91%
dice scores, sensitivity, and specificity, respectively. Saxena
et al. [44] utilized ResNet-50, Inception V3, VGG-16, and
transfer learning techniques to classify the images; the
ResNet-50 outperformed with 95% accuracy. In [27], Çinar
and Yildirim utilized numerous convolutional neural
network (CNN) models, such as Inception V3, GoogLeNet,
ResNet-50, AlexNet, and DenseNet-201, for classifying the
MRI datasets; whether they acquired respectable accuracies
in every case. Khwaldeh et al. [45] customized the AlexNet
architecture and employed it to classify brain MRIs, and
obtained 91% accuracy. Preethi and Aishwarya [46]
presented an architecture that combined wavelet-based
GLCM to develop matrices of features. They utilized a deep
neural network classifier to detect and categorize brain MRIs
and obtained 92% accuracy. In [28], Hemanth et al. pro-
posed a modified deep convolutional neural network
(DCNN) to classify brain tumors, and they obtained excel-
lent model performance in terms of accuracy. Khan et al.
[47] introduced a computerized multimodel categorization
technique to detect and classify brain tumor images. They
utilized two presupervised CNN models, known as VGG16
and VGG19, to extract features. They deployed the
BRATS2018 dataset to validate their model and obtained
about 97.80% accuracy.
Journal of Sensors 3

Input MR images

Training image set Test image set

Preprocessing Preprocessing

Feature extraction Feature extraction

Extracted features Extracted features

Classifier
Prediction model (BayesNet, SMO,
NB, RF)

Classified result

Glioma Meningioma No-tumor Pituitary

Figure 1: Block diagram of the proposed method for identification and classification of brain tumors.

accuracy was obtained. Updated versions of CNN architec-

Table 1: Statistical description of data.
tures, namely, AlexNet and VGG16, were introduced by
Image type Total images Training images Testing images Ali et al. [15] with 96% and 98% of accuracy, respectively.
Glioma 900 630 270
Deepak and Ameer [32] proposed a fully computerized
technique for classifying brain tumor images into three
Meningioma 900 630 270
categories: glioma, meningioma, and pituitary tumors.
No-tumor 900 630 270 Their designed architecture was an updated version of
Pituitary 900 630 270 the GoogLeNet model and obtained 98% accuracy. In
Total 3600 2520 1080 another study, Sultan et al. [34] presented a DCNN model
for accurately classifying MRI data into glioma, meningi-
oma, and pituitary tumors, where they achieved 96% accu-
racy. Abiwinanda et al. [35] designed numerous CNN
In [48], Siar and Teshnehlab utilized a central cluster- architectures for classifying brain tumors into three classes:
ing algorithm to extract features, then fed those features glioma, meningioma, and pituitary tumors. One of the
directly to the CNN model and achieved 96% accuracy. architectures showed the best result with 98.51% of accu-
Kuang et al. [49] introduced a 3D CNN model to classify racy. In [36], Seetha and Raja classified the brain tumor
the BRATS2018 brain tumor image dataset. Preprocessing images using the CNN. They extracted the features using
was performed using the grey-level normalization tech- ImageNet and obtained 97.50% accuracy.
nique and contrast adjustment. Afterward, those images Every method has some advantages and disadvantages.
were transferred to the 3D CNN model for classification. Some research gaps are identified in the above-mentioned
They achieved a dice score of 92%. In another study papers and summarized briefly as follows.
[29], Zhou et al. considered 3D images as 2D cross-
sections to utilize those images as the input dataset. They (i) Many researchers have used different techniques in
employed different models, such as DenseNet-RNN, Den- particular to determine the region of interest
seNet-LSTM, and DenseNet-DenseNet, and obtained accu- (ROI) for extracting features
racies of 87%, 91% and 92%, respectively. In [31],
Krishnammal and Raja utilized a presupervised AlexNet (ii) Many researchers have used statistical features and
model to classify the MRI dataset. Features were extracted texture features, but these have not given so much
through curvelet transform and GLCM matrix, and 100% good accuracy with respect to deep features
4 Journal of Sensors

Fully connected layer

Convolution + Convolution +
Output
ReLU Pooling ReLU Pooling
Glioma

Meningioma

No-tumor
•••••

Pituitary

Figure 2: Basic CNN architecture.

1st CON 11×11, 55×55×96

227×227×3 Stride = 4 55×55×96 2nd CON 5×5,
Max pool 3×3 3rd CON 3×3,
Kemels = 96 Padding = 2, 27×27×256 Max pool 3×3, Padding = 1,
Stride = 2 Kemels = 256 Stride = 2 13×13×256 Kemels = 384 13×13×384
55×55 55×55
227×227 {(227-11)/4}+1=55 27×27
{(55-3)/2}+1=27 ⁎
{(27+2 2-5)/1}+1=27 {(27-3)/2}+1=13 13×13 ⁎
13×13
{(13+2 1-3)/1}+1=13

{(13+2⁎1-3)/1}+1=13
4th CON 3×3,

Kemels = 384
Padding = 1,
1000 softmax Fully connected 8 13×13 13×13
Fully connected 7
6×6
Fully connected 6 {(13-3)/2}+1=6 ⁎
{(13+2 1-3)/1}+1=13

6×6×256 Max pool 3×3, 5th CON 3×3,

Stride = 2 13×13×256 Padding = 1, 13×13×384
Kemels = 256
4096 4096

Figure 3: The proposed architecture of AlexNet.

(iii) Most of the studies have used good quality MRI
images. Raw images or noisy images have been
found well dealt by deep learning methods
(iv) Manual classification of brain tumors from MRIs is
time-consuming, nonreproducible, and impractical
if the data size is larger
Thus, an excellent model is required for feature extraction
and classification of brain tumors accurately from MRIs.
rated in the upcoming subsections. We have used AlexNet
CNN in the MATLAB platform [50] for categorizing the
dataset as training and test as well as extracting the features
from it. Then, the extracted features are employed in the
WEKA platform [51] to classify the brain tumors of MRIs
as no-tumor, glioma, meningioma, and pituitary tumors uti-
lizing prominent BayesNet, SMO, NB, and RF classifiers and
establishing a prediction model. The performance of the
model is assessed by the classified results.

3. Materials and Methods
3.1. Research Implementation Block Diagram. Figure 1 shows
the overall working architecture of our proposed model to
accurately identify and classify brain tumors from MRIs.
Each key component of this architecture, such as the dataset
including training and test data, preprocessing, feature
extraction, prediction model, classified results, and perfor-
mance measurement of the diagram, are sequentially nar-
3.2. Data Description. The selection of an appropriate data-
set is the first and foremost concern for identifying and clas-
sifying medical images. Therefore, we collected the dataset
from a trustworthy website, the Kaggle database [52, 53], a
well-known web-based data source for brain MRIs. From
[52], we have utilized the glioma, meningioma, and pituitary
tumor data, and from [53], we have utilized the no tumor
data. In total, 3600 images have been used in this study,
where each class contains the same number of images (900
 Table 2: Confusion matrix for (a) test data itself, (b) CVF 4, (c) CVF 10, and (d) PS 34% of test data using BayesNet, SMO, NB, and RF classifiers.

(a)

BayesNet classifier SMO classifier

Glioma Meningioma No tumor Pituitary Glioma Meningioma No tumor Pituitary
Journal of Sensors

Glioma 203 36 6 25 Glioma 252 17 1 0

Meningioma 30 175 27 38 Meningioma 6 256 3 5
No-tumor 5 22 243 0 No-tumor 0 1 269 0
Pituitary 14 37 3 216 Pituitary 3 4 0 263
NB classifier RF classifier
Glioma Meningioma No tumor Pituitary Glioma Meningioma No tumor Pituitary
Glioma 197 32 6 35 Glioma 270 0 0 0
Meningioma 44 144 30 52 Meningioma 0 270 0 0
No-tumor 5 29 235 1 No-tumor 0 0 270 0
Pituitary 22 37 4 207 Pituitary 0 0 0 270

(b)

BayesNet classifier SMO classifier

Glioma Meningioma No tumor Pituitary Glioma Meningioma No tumor Pituitary
Glioma 181 51 9 29 Glioma 224 34 3 9
Meningioma 39 158 30 43 Meningioma 43 200 15 12
No-tumor 6 31 233 0 No-tumor 5 11 251 3
Pituitary 14 47 4 205 Pituitary 8 18 1 243
NB classifier RF classifier
Glioma Meningioma No tumor Pituitary Glioma Meningioma No tumor Pituitary
Glioma 191 35 8 36 Glioma 207 45 4 14
Meningioma 43 138 35 54 Meningioma 38 187 23 22
No-tumor 5 29 235 1 No-tumor 3 10 256 1
Pituitary 20 37 5 208 Pituitary 7 16 2 245

(c)

BayesNet classifier SMO classifier

Glioma Meningioma No tumor Pituitary Glioma Meningioma No tumor Pituitary
Glioma 183 50 7 30 Glioma 211 44 5 10
Meningioma 39 153 32 46 Meningioma 31 208 16 15
No-tumor 8 30 232 0 No-tumor 4 9 254 3
5
 6

Table 2: Continued.

Pituitary 20 41 3 206 Pituitary 9 15 2 244

NB classifier RF classifier
Glioma Meningioma No tumor Pituitary Glioma Meningioma No tumor Pituitary
Glioma 189 38 6 37 Glioma 214 36 5 15
Meningioma 48 134 33 55 Meningioma 38 182 26 24
No-tumor 5 29 235 1 No-tumor 4 9 256 1
Pituitary 22 39 4 205 Pituitary 9 14 5 242

(d)

BayesNet classifier SMO classifier

Glioma Meningioma No tumor Pituitary Glioma Meningioma No tumor Pituitary
Glioma 116 37 5 26 Glioma 132 38 4 10
Meningioma 25 98 27 28 Meningioma 34 116 16 12
No-tumor 6 17 152 0 No-tumor 3 9 161 2
Pituitary 10 25 1 140 Pituitary 7 18 1 150
NB classifier RF classifier
Glioma Meningioma No tumor Pituitary Glioma Meningioma No tumor Pituitary
Glioma 120 36 5 23 Glioma 134 35 5 10
Meningioma 31 84 29 34 Meningioma 21 107 23 27
No-tumor 6 15 154 0 No-tumor 2 9 163 1
Pituitary 9 21 2 144 Pituitary 2 17 1 156
Journal of Sensors
Journal of Sensors 7

1
1

0.8
0.8

True positive rate

True positive rate

0.6
0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
False positive rate False positive rate
BayesNet NB BayesNet NB
SMO RF SMO RF
(a) (b)
1 1

0.8 0.8
True positive rate
True positive rate

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

False positive rate False positive rate

BayesNet NB BayesNet NB
SMO RF SMO RF
(c) (d)

Figure 4: ROC curves of (a) glioma, (b) meningioma, (c) no tumor, and (d) pituitary tumors for test data itself using BayesNet, SMO, NB,
and RF classifiers.

per class). The elaborate allocation of the data collected and
deployed in this study is shown in Table 1. We split the data-
set (3600 MRIs) into training and test datasets: 2520 (70% of
the total dataset) images are utilized as training images, and
the rest 1080 (30% of the total dataset) images as test images,
using AlexNet CNN. Training data are deployed for model
prediction; on the other hand, test data are applied for testi-
fying the model’s performance.
3.3. Preprocessing. Preprocessing is a technique that aug-
ments the quality of images and makes them able for further
steps. In a certain cases, the separation between normal and
abnormal tissue is complicated because of the high-clamor
level. As a result, specialists may normally commit errors
in diagnosis. Then again, minor contrasts between normal
and abnormal tissues can likewise be concealed by the
clamor. Therefore, it is important to evacuate the conceiv-
able clamors through preprocessing. In addition, enhance-
ment of the visual nature of images offers tremendous
favor to the specialists. In this research, MRIs are prepro-
cessed with an anisotropic filter. The size of the images in
the dataset of MRIs is not the same. These different sizes
of images illustrate the input layer of the CNN network;
therefore, these images need to be preprocessed, standard-
ized, and resized to 227 × 227 pixels.
3.4. Feature Extraction. One presupervised CNN model,
AlexNet [15, 31], is utilized in this investigation to include
extraction, because it is a fantastically incredible model fit
for accomplishing high-level accuracy on testing datasets.
AlexNet is the most important known deep CNN structure
which comprises twenty-five layers, eight of which add to
8 Journal of Sensors

1 1
0.9 0.9
0.8 0.8
0.7 0.7

True positive rate

True positive rate

0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
False positive rate False positive rate
BayesNet NB BayesNet NB
SMO RF SMO RF
(a) (b)
1 1
0.9
0.8 0.8
0.7
True positive rate
True positive rate

0.6 0.6
0.5
0.4 0.4
0.3
0.2 0.2
0.1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
False positive rate False positive rate
BayesNet NB BayesNet NB
SMO RF SMO RF
(c) (d)

Figure 5: ROC curves of (a) glioma, (b) meningioma, (c) no tumor, and (d) pituitary tumors for CVF 4 of the test data using BayesNet,
SMO, NB, and RF classifiers.

learning by modifying loads. Five of these are known as con-
volution layers while staying three are known as fully con-
nected (FC) layers. In the AlexNet design, max-pooling
layers are connected to convolution layers in series with each
other. The first convolution layer and then max-pooling
layers utilize fluctuating kernel sizes [54]. The layer of
max-pooling ensues the layer of convolution [54, 55]. FC6
is the first FC layer, and FC7 is the second FC layer in acti-
vation are utilized to remove the features of vectors. There
are altogether 4096 features of FC6 and FC7 vectors in the
AlexNet CNN design [56].
3.5. Prediction Model. Consequently, we have developed pre-
dictive models based on features extracted from the trained
data. A prediction model is a procedure through which
future results or conducts are predicted dependent on the
former and present data. It is a measurable examination pro-
cedure that empowers the assessment and count of the like-
lihood of specific outcomes identified with programming or
frameworks. The prediction model works by gathering
information, making a factual model, and applying probabi-
listic methods to anticipate the possible results.
3.6. Proposed Network Architecture
3.6.1. Convolutional Neural Network. The convolutional
neural network (CNN) is a sort of multilayered feed-
forward neural network and one of the significant ideas of
the deep learning method. Its basic structure consists of sev-
eral convolution layers, a rectified linear unit (ReLU) layer,
Journal of Sensors 9

1 1

0.8 0.8

True positive rate

True positive rate

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
False positive rate False positive rate
BayesNet NB BayesNet NB
SMO RF SMO RF
(a) (b)
1 1

0.8 0.8
True positive rate
True positive rate

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
False positive rate False positive rate
BayesNet NB BayesNet NB
SMO RF SMO RF
(c) (d)

Figure 6: ROC curves of (a) glioma, (b) meningioma, (c) no tumor, and (d) pituitary tumors for the CVF 10 of test data using BayesNet,
SMO, NB and RF classifiers.

pooling layers, and FC layers, as shown in Figure 2. The
basic function of the convolution layer is to extract the
local features of the existing layers. The ReLU layer per-
forms element-by-element activation. The pooling layers
are introduced for downsampling. Max-pooling is com-
monly utilized to reduce the accuracy of the feature maps
by incorporating linguistically similar features. To over-
come the fitting issues, some neurons have been removed
from the CNN architecture, known as dropouts. At the
end, the FC layer yields the class score value ranging from
0 to 1 which is deployed to attain the classification deci-
sions. Essentially, SoftMax layers are introduced by default
in this motive.
3.6.2. AlexNet. Figure 3 shows the architecture of an AlexNet
which is mainly comprised of five convolutional layers
(11 × 11, 5 × 5, 3 × 3, 3 × 3, 3 × 3), three max-pooling layers
(3 × 3), and three FC layers. The first two max-pooling
layers are deployed after the first two convolutional layers
sequentially. The third, fourth, and fifth convolutional
layers are connected directly. The third max-pooling layer
is inserted after the fifth convolutional layer, and its out-
put is fed into a series of three FC layers. The success of
AlexNet is usually credited to ReLU, stochastic gradient
descent (SGD), dropout, etc. ReLU is introduced for accel-
erating the speed of training processes. The value of the
convolutional kernel is extracted by optimizing the total
cost function and applying the SGD algorithm. To elimi-
nate the overfitting problem, a dropout layer is applied
in the first two FC layers. The third FC layer, also known
as the softmax layer, is used to classify various objects in a
CNN [57].
10 Journal of Sensors

1 1

0.8 0.8

True positive rate

True positive rate

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
False positive rate False positive rate
BayesNet NB BayesNet NB
SMO RF SMO RF
(a) (b)
1 1
0.9
0.8 0.8
0.7
True positive rate
True positive rate

0.6 0.6
0.5
0.4 0.4
0.3
0.2 0.2
0.1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
False positive rate False positive rate
BayesNet NB BayesNet NB
SMO RF SMO RF
(c) (d)

Figure 7: ROC curves of (a) glioma, (b) meningioma, (c) no tumor, and (d) pituitary tumors for PS 34% of the test data using BayesNet,
SMO, NB, and RF classifiers.

3.7. Classifiers be materialized by training the SVMs. At this issue, the

3.7.1. BayesNet Classifier. The BayesNet classifier is a kind of
directed acyclic graph which enciphers a joint probability
distribution over a set of random variables [58, 59]. It is a
twosome of B = ðG, θÞ, where G represents the graph whose
vertices correspond to the random variables X1… Xn, as well
as whose edges denote direct dependencies between the var-
iables. θ denotes the set of parameters which quantifies the
network. For the BayesNet classification model, in the pres-
ent study, a simple estimator is used, and the batch size is
chosen as 100 to get better classification accuracy.
3.7.2. Sequential Minimal Optimization Classifier. Sequential
minimal optimization (SMO) is an important algorithm that
can solve the quadratic programming problem as well as
least plausible optimization problem includes the Lagrange
multipliers which have two numbers, because the Lagrange
multipliers must comply with a direct equity limitation.
First, it discovers two multipliers, and then it attempts to
enhance them. It will be repeated until it meets the equi-
librium state [60]. Penalty parameter (C) is chosen as 1,
and s polynomial kernel is selected based on the trial-
and-error method for the SMO classification model,
because they provide better model performance.
3.7.3. Naïve Bayes Classifier. The Naïve Bayes (NB) classi-
fier is a probabilistic machine learning model based on
Bayes theorem; that is, very useful for classifying a large
amount of data. It anticipates based on the probability of
an object. Thus, the model is developed, and the common
Journal of Sensors 11

Table 3: Calculated values of ROC and PRC areas for test data itself, CVF 4, CVF 10, and PS 34% of test data using BayesNet, SMO, NB, and
RF classifiers.

ROC areas PRC areas

Testing options
BayesNet SMO NB RF BayesNet SMO NB RF
Test data itself 0.933 0.984 0.900 1.000 0.798 0.944 0.716 1.000
CVF 4 0.899 0.925 0.890 0.956 0.723 0.795 0.697 0.894
CVF 10 0.901 0.924 0.891 0.956 0.727 0.791 0.697 0.897
PS 34% 0.895 0.889 0.892 0.932 0.718 0.713 0.703 0.848

1 1
Normalized expected cost

Normalized expected cost

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Probability cost function Probability cost function

BayesNet NB BayesNet NB
SMO RF SMO RF
(a) (b)
1 1
Normalized expected cost

Normalized expected cost

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Probability cost function Probability cost function
BayesNet NB BayesNet NB
SMO RF SMO RF
(c) (d)

Figure 8: Cost curves of (a) glioma, (b) meningioma, (c) no tumor, and (d) pituitary tumors for the test data itself using BayesNet, SMO,
NB, and RF classifiers.

input to the model is batch size (number of experiments
carried out) [61], whose value is selected as 100 for better
accuracy of the model.
3.7.4. Random Forest Classifier. The random forest (RF)
classifier is a supervised machine learning algorithm that
creates a forest with a bunch of trees. Its working proce-
dure consists of building simple decision trees and finally
making the decision based on the votes. It looks for the
best feature in every step (splitting nodes and growing
trees), which predominantly results in a superior model.
Batch size, number of features, and seed value are the com-
mon inputs in the RF model [61]. For better performance,
we have considered batch size as 100 and seed value as 1.
12 Journal of Sensors

1 1

Normalized expected cost

Normalized expected cost

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Probability cost function Probability cost function
BayesNet NB BayesNet NB
SMO RF SMO RF
(a) (b)
1 1
Normalized expected cost

Normalized expected cost

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Probability cost function Probability cost function
BayesNet NB BayesNet NB
SMO RF SMO RF
(c) (d)

Figure 9: Cost curves of (a) glioma, (b) meningioma, (c) no tumor, and (d) pituitary tumors for CVF 4 of the test data using BayesNet,
SMO, NB, and RF classifiers.

3.8. Classified Result. In this research, it is expected that the the CVF 10 [19], and percentage split (PS) 34% [51] of
final classified results will be obtained as glioma, meningi- test data, by employing the Equations (1)–(11) to measure
oma, no-tumor, and pituitary tumors from MRI datasets. the performance of our proposed model.
Then, we wish that the qualitative and quantitative values
of the performance measurement of the proposed model’s TP + TN
will be attained to evaluate the model performance. Accuracy = , ð1Þ
TP + TN + FP + FN

3.9. Performance Measurement. A confusion matrix has

TP
been utilized to calculate the performance of a classifier. Sensitivity or Recall or True Positive Rate = , ð2Þ
The performances are described by four components of TP + FN
the confusion matrix, such as true positive (TP), true neg-
ative (TN), false positive (FP), and false negative (FN), TN
Specificity or True Negative Rate = , ð3Þ
which are whole in number and not in rate. The values TN + FP
of the performance parameters, such as accuracy, sensitiv-
ity, specificity, kappa statistics, false positive rate, preci- Observed Agreement − Expected Agreement
sion, f-measure, and Matthew’s correlation coefficient Kappa statistics = :
1 − Expected Agreement
(MCC), have been calculated for four testing options: the
test data itself, the cross-validation fold (CVF) 4 [16, 34], ð4Þ
Journal of Sensors 13

1 1

0.8 0.8
Normalized expected cost

Normalized expected cost

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Probability cost function Probability cost function
BayesNet NB BayesNet NB
SMO RF SMO RF
(a) (b)
1 1

0.8 0.8
Normalized expected cost

Normalized expected cost

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Probability cost function Probability cost function
BayesNet NB BayesNet NB
SMO RF SMO RF
(c) (d)

Figure 10: Cost curves of (a) glioma, (b) meningioma, (c) no tumor, and (d) pituitary tumors for CVF 10 of the test data using BayesNet,
SMO, NB, and RF classifiers.

Here, 4 marginal, r1m is the row 1 marginal, r2m is the row 2 mar-
ginal, r3m is the row 3 marginal, r4m is the row 4 marginal,
Total Number of Correctly Classified and n is the total number of observations.
Observed Agreement = ,
n
ð5Þ FP
False Positive Rate ðFPRÞ = , ð7Þ
FP + TN
and
FN
Expected Agreement False Negative RateðFNRÞ = , ð8Þ
ÀÀ 1 Á Á ÀÀ Á Á ÀÀ Á Á ÀÀ Á Á FN + TP
cm × r1m /n + c2m × r2m /n + c3m × r3m /n + c4m × r4m /n
= ,
n TP
Precision = , ð9Þ
ð6Þ TP + FP

where c1m is the column 1 marginal, c2m is the column 2 Precision × Recall
F − measure = 2 × , ð10Þ
marginal, c3m is the column 3 marginal, c4m is the column Precision + Recall
14 Journal of Sensors

1 1

0.8 0.8
Normalized expected cost

Normalized expected cost

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

Probability cost function Probability cost function

BayesNet NB BayesNet NB
SMO RF SMO RF
(a) (b)
1 1

0.8 0.8
Normalized expected cost

Normalized expected cost

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

Probability cost function Probability cost function

BayesNet NB BayesNet NB
SMO RF SMO RF
(c) (d)

Figure 11: Cost curves of (a) glioma, (b) meningioma, (c) no tumor, and (d) pituitary tumors for PS 34% of the test data using BayesNet,
SMO, NB, and RF classifiers.

that represents the performance of the classifier based on

Matthew’ s correlation coefficient ðMCCÞ
TP × TN − FP × FN
= pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi :
ðTP + FPÞðTP + FNÞðTN + FPÞðTN + FNÞ
Different types of curves, such as the receiver operating
characteristics (ROC) curve, precision-recall curve (PRC),
and cost curve, have been plotted to understand the per-
formance of every classifier. The ROC curve is utilized to
quantify how well the zone could be recognized from
noise by plotting the true positive rate against the false
positive rate [28, 32]. The PRC is a well-known model
execution measurement to assess the binary classification
model [62]. A cost curve can be defined as a visual curve
the misclassification cost [63].
ð11Þ
4. Results and Discussion
4.1. Confusion Matrix. Table 2 represents the confusion
matrix for the test data itself: CVF 4, CVF 10, and PS 34%
of the test data using BayesNet, SMO, NB, and RF classifiers,
respectively. For the test data itself, the table shows that our
models with the BayesNet classifier can classify 203 gliomas,
175 meningiomas, 243 no-tumors, and 216 pituitary tumor
images accurately. The SMO classifier can classify 252 glio-
mas, 256 meningiomas, 269 no-tumors, and 263 pituitary
images accurately. The NB classifier can classify 197 glioma
cases, 144 meningioma cases, 235 no-tumor cases, and 207
Journal of Sensors 15

Table 4: Calculated values of different performance metrics for test data itself, CVF 4, CVF 10, and PS 34% of test data using BayesNet,
SMO, NB, and RF classifiers.

Test data itself CVF 4 of the test data

Performance metrics
BayesNet SMO NB RF BayesNet SMO NB RF
Accuracy (%) 88.750 98.148 86.250 100 85.972 92.500 85.741 91.435
Sensitivity (%) 77.500 96.296 72.500 100 71.946 85.000 71.482 82.871
Specificity (%) 92.500 98.763 90.834 100 90.648 95.000 90.494 94.290
FPR 0.075 0.012 0.092 0.000 0.094 0.050 0.095 0.057
FNR 0.225 0.037 0.275 0.000 0.281 0.150 0.286 0.171
Precision 0.775 0.963 0.722 1.000 0.722 0.850 0.710 0.826
F-measure 0.775 0.963 0.722 1.000 0.720 0.850 0.711 0.827
Kappa 0.700 0.950 0.633 1.000 0.626 0.800 0.620 0.772
MCC 0.700 0.951 0.632 1.000 0.627 0.800 0.618 0.770
CVF 10 of the test data PS 34% of the test data
BayesNet SMO NB RF Bayes net SMO NB RF
Accuracy (%) 85.834 92.454 85.324 91.389 85.484 89.200 85.204 89.271
Sensitivity (%) 71.667 84.907 70.648 82.778 71.102 78.534 70.557 78.680
Specificity (%) 90.556 94.969 90.216 94.259 90.324 92.790 90.135 92.851
FPR 0.094 0.050 0.098 0.057 0.097 0.072 0.099 0.071
FNR 0.283 0.151 0.294 0.172 0.289 0.215 0.295 0.214
Precision 0.717 0.849 0.702 0.824 0.709 0.783 0.697 0.783
F-measure 0.716 0.849 0.703 0.825 0.707 0.784 0.698 0.783
Kappa 0.622 0.799 0.609 0.770 0.613 0.712 0.606 0.714
MCC 0.622 0.799 0.607 0.769 0.612 0.711 0.602 0.713

pituitary cases accurately. The RF classifier, our model, can
classify all the classes accurately.
For CVF 4 of the test data, in the table, we can see that
our model with the BayesNet classifier can classify 181 gli-
oma images, 158 meningioma images, 233 no-tumors, and
205 pituitary images accurately; with the SMO classifier, it
can classify 224 glioma images, 200 meningioma images,
251 no-tumors, and 243 pituitary images accurately; with
the NB classifier, it can classify 191 glioma images, 138
meningioma images, 235 no-tumors, and 208 pituitary
images accurately; and with the RF classifier can classify
207 glioma images, 187 meningioma images, 256 no-tumors,
and 245 pituitary images accurately.
As shown in the table, for CVF 10 of the test data, while
we utilizing BayesNet classifier, our model can classify 183
glioma images, 153 meningioma images, 232 no-tumors,
and 206 pituitary images accurately; while utilizing SMO
classifier, our model can classify 211 glioma images, 208
meningioma images, 254 no-tumors, and 244 pituitary
images accurately; again, while utilizing NB classifier, our
model can classify 189 glioma images, 134 meningioma
images, 235 no-tumors, and 205 pituitary images accurately;
while utilizing RF classifier, our model can classify 214 gli-
oma images, 182 meningioma images, 256 no-tumors, and
242 pituitary images accurately.
In the case of PS 34% of the test data, the table dem-
onstrates that while we have utilized BayesNet classifier,
our model can classify 116 glioma images, 98 meningioma
images, 152 no-tumors, and 140 pituitary images accu-
rately; while we have utilized SMO classifier, our model
can classify 132 glioma images, 116 meningioma images,
161 no-tumors, and 150 pituitary images accurately; again,
while we have utilized NB classifier, our model can classify
120 glioma images, 84 meningioma images, 154 no-
tumors, and 144 pituitary images accurately; while we have
utilized RF classifier, our model can classify 134 glioma
images, 107 meningioma images, 163 no-tumors, and 156
pituitary images accurately.
4.2. ROC and PRC Curves of Test Data. We have plotted
ROC (Figures 4–7) and PRC (not shown in this paper)
curves of glioma, meningioma, no-tumor, and pituitary
tumors for the test data itself, CVF 4, CVF 10, and PS 34%
of test data using BayesNet, SMO, NB, and RF classifiers,
respectively. And then, we have calculated the values of the
area under each ROC and PRC curves. The average values
are summarized in Table 3. As shown in the table, the aver-
age values of the ROC area (i) for the test data itself are
0.933, 0.984, 0.900, and 1.000; (ii) for CVF 4 are 0.899,
0.925, 0.890, and 0.956; (iii) for CVF 10 are 0.901, 0.924,
0.891, and 0.956; and (iv) for PS 34% are 0.895, 0.889,
0.892, and 0.932 of the test data using BayesNet, SMO, NB,
and RF classifiers, respectively. These values indicate that
the RF classifier performs better than the other classifiers
in every case. The table also similarly reveals that the RF
classifier gives the highest PRC area values of 1.000, 0.894,
0.897, and 0.848 for the test data itself, CVF 4, CVF 10,
and PS 34%, respectively.
16 Journal of Sensors

Table 5: Comparison with other existing findings.

Ref. Classifiers Database Accuracy (%)

AlexNet 60.20
[15] BRATS2015
VGG16 98.20
[26] CNN Nanfang Hospital, General Hospital, China 97.3
[27] CNN Kaggle 97.2
DCNN 94.5
[28] M/s. Devaki Scan Centre, India
MDCNN 96.4
DenseNet+RNN 84.61
Nanfang Hospital, General Hospital,
[29] DenseNet+LSTM 92.13
Tianjin Medical University, China
DenseNet+DenseNet 86.68
[30] CNN BRATS 97.87
[32] GoogLeNET FIGSHARE 98
Nanfang Hospital and General Hospital,
96.13
[34] CNN Tianjin Medical University, China
REMBRANDT 98.7
[35] CNN FIGSHARE 98.51
[36] CNN BRATS2015, radiopedia 97.5
[64] CNN FIGSHARE 98.71
BRATS2012, BRATS2013,
[65] Deep wavelet auto-encoder 99.30
BRATS2014, BRATS2015
[66] Deep recurrent long short-term memory BRATS2015 99
[67] GoogLeNet+AlexNet TJU Hospital, China. 99.51
AlexNet CNN+BayesNet 88.75
AlexNet CNN+SMO 98.15
[proposed] Kaggle
AlexNet CNN+NB 86.25
AlexNet CNN+RF 100

4.3. Cost Curve Analysis of Test Data. We have also analyzed
cost curves which yields model performance based on the
minimum misclassification cost. The minimum misclassifica-
tion cost can also be defined as the minimum area under the
bottom of the envelope. Therefore, the minimum region below
the lower end of the range lowers the cost of misclassification
and thus provides better classification performance. Figures 8–
11 represent the cost curves for different testing options, such
as the test data itself, CVF 4, CVF 10, and PS 34% of the test
data employing BayesNet, SMO, NB, and RF classifiers for gli-
oma, meningioma, no-tumor, and pituitary classes, respec-
tively. As seen in Figure 8, the RF classifier very clearly offers
minimal misclassification cost; that is, better classification per-
formance. However, the SMO classifier shows a slightly lower
misclassification cost in Figures 9–11.
4.4. Performance Evaluation. Finally, we have evaluated the
performance of our models based on their respective confusion
matrices. We have calculated the values of accuracy, sensitivity,
specificity, FPR, FNR, precision, f-measure, kappa and MCC
and summarized them in Table 4, which represents the values
for the test data itself: CVF 4, CVF 10, and PS 34%, respectively.
For the test data itself, Table 4 shows that the values of accuracy,
sensitivity, specificity, FPR, FNR, precision, f-measure, kappa,
and MCC are higher for the RF classifier compared to those
values of the other classifiers, exhibiting that the RF classifier
gives excellent performance with 100% accuracy.
For the case of CVF 4 of test data, Table 4 shows that the
values of accuracy, sensitivity, specificity, precision, f-mea-
sure, kappa, and MCC are higher and the values of FPR
and FNR are lower for the SMO classifier than those values
for the other classifiers, indicating that the SMO classifier
gives better performance with 92.50% accuracy. Similar
results are seen in the case of CVF 10 of test data, where
the SMO classifier gives 92.45% accuracy.
However, PS 34% of the test data in the table shows that
the values of accuracy, sensitivity, specificity, FPR, FNR,
kappa, and MCC are higher for the RF classifier than the
other classifiers. The values of precision are the same for
both SMO and RF classifiers, and the values of f-measure
are very slightly lower for the RF classifier compared to the
SMO classifier. From these statistical values, we can assume
that the RF classifier reveals a better performance with
89.27% accuracy.
Journal of Sensors
4.5. Comparison with Other Works. In this study, we have
obtained the highest accuracy of 88.75%, 98.15%, 86.25%,
and 100% for AlexNet CNN with (i) BayesNet, (ii) NB,
(iii) SMO, and (iv) RF classifiers using the test data itself,
respectively. Table 5 represents the comparison of our pro-
posed model with several contemporary findings in terms
of accuracy. As shown in the table, our model is highly
appreciable, effective, and novel.
5. Conclusion
In this study, a novel and effective technique for feature
extraction and classification of brain MRIs is presented
for classifying brain tumors into glioma, meningioma, no-
tumor, and pituitary classes. An AlexNet CNN with Bayes-
Net/SMO/NB/RF models is approached. Performances of
the model are evaluated using the test dataset. We have
calculated the values of performance parameters from
respective confusion matrix, namely, accuracy, sensitivity,
specificity, FPR, FNR, precision, f-measure, kappa statistics,
MCC, ROC area, and PRC area for four testing options:
the test data itself, CVF 4, CVF 10, and PS 34% of test
data. We have also explained ROC, PRC, and cost curves
to measure the classification performance. We have
acquired 100%, 98.15%, 88.75%, and 86.25% of accuracy
using the AlexNet CNN+RF, AlexNet CNN+SMO, AlexNet
CNN+BayesNet, and AlexNet CNN + NB models, respec-
tively. This study presents a paramount technique for the
classification of brain tumors from MRIs.
However, the proposed model was evaluated using a
moderately sized dataset, which is one of the limitations of
the current work. So, it is essential to evaluate the model
with big data size in the future to see how well it performs.
Another limitation is that the proposed approach has not
been tested employing the MRIs of real-patient in Bangla-
desh. In the future, we will try to deploy this work on a
real-time medical diagnostic system by acquiring the MRIs
from different hospitals and diagnostic centers in Bangla-
desh. We believe that our approach is very much applicable
in the health sector, offering innovative solutions with high
precision for medical imaging, especially for diagnosing
brain tumors more accurately.
Data Availability
Experimental codes and the data of this work are available
from the corresponding author upon request at any time.
Conflicts of Interest
The authors declare no conflicts of interest.
Acknowledgments
We wish to express our gratitude to Khulna University,
Khulna, Bangladesh-9208.
17
References
[1] D. N. Louis, A. Perry, G. Reifenberger et al., “The 2016
World Health Organization classification of tumors of the
central nervous system: a summary,” Acta Neuropathologica,
vol. 131, no. 6, pp. 803–820, 2016.
[2] N. B. Bahadure, A. K. Ray, and H. P. Thethi, “Image analysis
for MRI based brain tumor detection and feature extraction
using biologically inspired BWT and SVM,” International
Journal of Biomedical Imaging, vol. 2017, Article ID 9749108,
12 pages, 2017.
[3] K. Popuri, D. Cobzas, A. Murtha, and M. Jägersand, “3D var-
iational brain tumor segmentation using Dirichlet priors on a
clustered feature set,” International Journal of Computer
Assisted Radiology and Surgery, vol. 7, no. 4, pp. 493–506,
2012.
[4] K. Fukushima, “Neocognitron: A self-organizing neural net-
work model for a mechanism of pattern recognition unaffected
by shift in position,” Biological Cybernetics, vol. 36, no. 4,
pp. 193–202, 1980.
[5] M. J. Mckeown and D. A. Ramsay, “Classification of astrocy-
tomas and malignant astrocytomas by principal components
Analysis and a neural net,” Journal of Neuropathology and
Experimental Neurology, vol. 55, no. 12, pp. 1238–1245,
1996.
[6] S. W. Coons, P. C. Johnson, B. W. Scheithauer, A. J. Yates, and
D. K. Pearl, “Improving diagnostic accuracy and interobserver
concordance in the classification and grading of primary glio-
mas,” American Cancer Society, vol. 79, no. 7, pp. 1381–1393,
1997.
[7] N. Belacela and M. R. Boulassel, “Multicriteria fuzzy assign-
ment method: a useful tool to assist medical diagnosis,” Artifi-
cial Intelligence in Medicine, vol. 21, no. 1-3, pp. 201–207,
2001.
[8] H. Gemmeke and N. V. Ruiter, “3D ultrasound computer
tomography for medical imaging,” Nuclear Instruments and
Methods in Physics Research, vol. 580, no. 2, pp. 1057–1065,
2007.
[9] E. S. A. El-Dahshan, H. M. Mohsen, K. Revett, and A. B. M.
Salem, “Computer-aided diagnosis of human brain tumor
through MRI: A survey and a new algorithm,” Expert Systems
with Applications, vol. 41, no. 11, pp. 5526–5545, 2014.
[10] B. Dogan, O. Demir, and S. K. Calik, “Computer-aided detec-
tion of brain tumors using morphological reconstruction,”
Uludag University Journal of The Faculty of Engineering,
vol. 21, no. 2, pp. 257–268, 2016.
[11] Y. Yan, X. J. Yao, S. H. Wang, and Y. D. Zhang, “A Survey of
computer-aided tumor diagnosis based on convolutional neu-
ral network,” Biology, vol. 10, no. 11, pp. 1–44, 2021.
[12] A. Rehman, M. A. Khan, T. Saba, Z. Mehmood, U. Tariq, and
N. Ayesha, “Microscopic brain tumor detection and classifica-
tion using 3D CNN and feature selection architecture,”
Microscopy Research and Technique, vol. 84, no. 1, pp. 133–
149, 2021.
[13] J. Amin, M. Sharif, A. Haldorai, M. Yasmin, and R. S. Nayak,
“Brain tumor detection and classification using machine learn-
ing: a comprehensive survey,” Complex & Intelligent Systems,
vol. 8, no. 4, pp. 3161–3183, 2022.
[14] A. A. Akinyelu, F. Zaccagna, J. T. Grist, M. Castelli, and
L. Rundo, “Brain tumor diagnosis using machine learning,
convolutional neural networks, capsule neural networks and
18
vision transformers, applied to MRI: a survey,” Journal Imag-
ing, vol. 8, no. 8, pp. 205–240, 2022.
[15] N. A. Ali, A. R. Syafeeza, L. J. Geok, Y. C. Wong, N. A. Hamid,
and A. S. Jaafar, “Design of automated computer-aided classi-
fication of brain tumor using deep learning,” in Intelligent and
Interactive Computing. Lecture Notes in Networks and Systems,
vol 67, V. Piuri, V. Balas, S. Borah, and S. Syed Ahmad, Eds.,
Springer, Singapore, 2019.
[16] B. S. Kumar, “Identification and classification of brain tumor
images using efficient classifier,” International Journal of Engi-
neering and Advanced Technology, vol. 8, no. 6, pp. 3677–3683,
2019.
[17] T. Ateeq, M. N. Majeed, S. M. Anwar et al., “Ensemble-classi-
fiers-assisted detection of cerebral microbleeds in brain MRI,”
Computers and Electrical Engineering, vol. 69, pp. 768–781,
2018.
[18] J. Cheng, W. Huang, S. Cao et al., “Enhanced performance of
brain tumor classification via tumor region augmentation
and partition,” PLoS One, vol. 10, no. 10, pp. 1–13, 2015.
[19] D. R. Patel, H. Thakker, M. B. Kiran, and V. Vakhariya, “Sur-
face roughness prediction of machined components using gray
level co-occurrence matrix and bagging tree,” FME Transac-
tions, vol. 48, no. 2, pp. 468–475, 2020.
[20] F. A. Zadeh, M. V. Ardalani, A. R. Salehi, R. J. Farahani,
M. Hashemi, and A. H. Mohammed, “An analysis of new fea-
ture extraction methods based on machine learning methods
for classification radiological images,” Computational Intelli-
gence and Neuroscience, vol. 2022, Article ID 3035426, 13
pages, 2022.
[21] G. T. Reddy, S. Bhattacharya, and S. S. Ramakrishnan, “An
ensemble based machine learning model for diabetic retinopa-
thy classification,” in 2020 International Conference on Emerg-
ing Trends in Information Technology and Engineering, pp. 1–
6, Vellore, India, 2020.
[22] R. H. Jhaveri, A. Revathi, K. Ramana, R. Raut, and R. K. Dha-
naraj, “A review on machine learning strategies for real-world
engineering applications,” Mobile Information Systems,
vol. 2022, Article ID 1833507, 26 pages, 2022.
[23] G. T. Reddy, M. P. K. Reddy, K. Lakshmanna, D. S. Rajput,
R. Kaluri, and G. Srivastava, “Hybrid genetic algorithm and a
fuzzy logic classifier for heart disease diagnosis,” Evolutionary
Intelligence, vol. 13, no. 2, pp. 185–196, 2020.
[24] U. Ahmed, R. H. Jhaveri, G. Srivastava, and J. C. W. Lin,
Explainable deep attention active learning for sentimental ana-
lytics of mental disorder, Association for Computing Machin-
ery, 2022.
[25] P. John, “Brain tumor classification using wavelet and texture
based neural network,” International Journal of Scientific and
Engineering Research, vol. 3, no. 10, pp. 1–7, 2012.
[26] F. J. Díaz-Pernas, M. Martínez-Zarzuela, M. Antón-Rodríguez,
and D. González-Ortega, “A deep learning approach for brain
tumor classification and segmentation using a multiscale convo-
lutional neural network,” Healthcare, vol. 9, no. 2, p. 153, 2021.
[27] A. Çinar and M. Yildirim, “Detection of tumors on brain MRI
images using the hybrid convolutional neural network archi-
tecture,” Medical Hypotheses, vol. 139, pp. 109684–109688,
2020.
[28] D. J. Hemanth, J. Anitha, A. Naaji, O. Geman, D. E. A.
Popescu, and L. Hoang Son, “A Modified deep convolutional
neural network for abnormal brain image classification,” IEEE
Access, vol. 7, pp. 4275–4283, 2019.
Journal of Sensors
[29] Y. Zhou, Z. Li, H. Zhu et al., “Holistic brain tumor screening
and classification based on DenseNet and recurrent neural
network,” in Brainlesion: Glioma, Multiple Sclerosis, Stroke
and Traumatic Brain Injuries. BrainLes 2018. Lecture Notes
in Computer Science, vol 11383Springer, Cham.
[30] T. Hossain, F. S. Shishir, M. Ashraf, M. A. Al Nasim, and F. M.
Shah, “Brain tumor detection using convolutional neural net-
work,” in 2019 1st International Conference on Advances in
Science, Engineering and Robotics Technology (ICASERT),
pp. 1–6, Dhaka, Bangladesh, 2019.
[31] P. M. Krishnammal and S. S. Raja, “Convolutional neural net-
work based image classification and detection of abnormalities
in MRI brain images,” in 2019 International Conference on
Communication and Signal Processing (ICCSP), pp. 0548–
0553, Chennai, India, 2019.
[32] S. Deepak and P. M. Ameer, “Brain tumor classification using
deep CNN features via transfer learning,” Computers in Biol-
ogy and Medicine, vol. 111, article 103345, pp. 1–12, 2019.
[33] E. I. Papageorgiou, P. P. Spyridonos, D. T. Glotsos et al., “Brain
tumor characterization using the soft computing technique of
fuzzy cognitive maps,” Applied Soft Computing, vol. 8, no. 1,
pp. 820–828, 2008.
[34] H. H. Sultan, N. M. Salem, and W. Al-Atabany, “Multi-classi-
fication of brain tumor images using deep neural network,”
IEEE Access, vol. 7, pp. 69215–69225, 2019.
[35] N. Abiwinanda, M. Hanif, S. T. Hesaputra, A. Handayani, and
T. R. Mengko, “Brain tumor classification using convolutional
neural network,” in World Congress on Medical Physics and
Biomedical Engineering 2018, pp. 183–189, Springer, 2019.
[36] J. Seetha and S. S. Raja, “Brain tumor classification using con-
volutional neural networks,” Biomedical & Pharmacology
Journal, vol. 11, no. 3, pp. 1457–1461, 2018.
[37] H. Selvaraj, S. T. Selvi, D. Selvathi, and L. Gewali, “Brain MRI
slices classification using least squares support vector
machine,” International Journal of Intelligent Computing in
Medical Sciences & Image Processing, vol. 1, no. 1, pp. 21–33,
2007.
[38] A. Sarkar, M. Maniruzzaman, M. S. Ahsan, M. Ahmad,
M. I. Kadir, and S. T. Islam, “Identification and classifica-
tion of brain tumor from MRI with feature extraction by
support vector machine,” in 2020 International Conference
for Emerging Technology (INCET), pp. 1–4, Belgaum, India,
2020.
[39] Z. Ullah, M. U. Farooq, S. H. Lee, and D. An, “A hybrid
image enhancement based brain MRI images classification
technique,” Medical Hypotheses, vol. 143, article 109922,
2020.
[40] A. Kharrat, K. Gasmi, M. B. Messaoud, N. Benamrane, and
M. Abid, “A hybrid approach for automatic classification of
brain MRI using genetic algorithm and support vector
machine,” Leonardo Journal of Sciences, vol. 17, no. 1,
pp. 71–82, 2010.
[41] N. V. Shree and T. N. R. Kumar, “Identification and classifica-
tion of brain tumor MRI images with feature extraction using
DWT and probabilistic neural network,” Brain Informatics,
vol. 5, no. 1, pp. 23–30, 2018.
[42] M. Arunachalam and S. Royappan Savarimuthu, “An efficient
and automatic glioblastoma brain tumor detection using shift-
invariant shearlet transform and neural networks,” Interna-
tional Journal of Imaging Systems and Technology, vol. 27,
no. 3, pp. 216–226, 2017.
Journal of Sensors 19

[43] S. Sajid, S. Hussain, and A. Sarwar, “Brain tumor detection and national Journal of Applied Engineering Research, vol. 55,
segmentation in MR images using deep learning,” Arabian pp. 1045–1049, 2015.
Journal for Science and Engineering, vol. 44, no. 11, [61] V. Vakharia and R. Gujar, “Prediction of compressive strength
pp. 9249–9261, 2019. and portland cement composition using cross-validation and
[44] P. Saxena, A. Maheshwari, and S. Maheshwari, “Predictive feature ranking techniques,” Construction and Building Mate-
modeling of brain tumor: a deep learning approach,” in Inno- rials, vol. 225, pp. 292–301, 2019.
vations in Computational Intelligence and Computer Vision. [62] T. Saito and M. Rehmsmeier, “The precision-recall plot is
Advances in Intelligent Systems and Computing, vol more informative than the ROC plot when evaluating binary
1189Springer, Singapore. classifiers on imbalanced datasets,” PLoS One, vol. 10, no. 3,
[45] S. Khawaldeh, U. Pervaiz, A. Rafiq, and R. S. Alkhawaldeh, pp. 1–21, 2015.
“Noninvasive grading of glioma tumor using magnetic reso- [63] C. Drummond and R. C. Holte, “Cost curves: An improved
nance imaging with convolutional neural networks,” Applied method for visualizing classifier performance,” Machine
Sciences, vol. 8, no. 1, p. 27, 2018. Learning, vol. 65, no. 1, pp. 95–130, 2006.
[46] S. Preethi and P. Aishwarya, “Combining wavelet texture fea- [64] R. Chelghoum, A. Ikhlef, A. Hameurlaine, and S. Jacquir,
tures and deep neural network for tumor detection and seg- “Transfer learning using convolutional neural network archi-
mentation over MRI,” Journal of Intelligent Systems, vol. 28, tectures for brain tumor classification from MRI images,” in
no. 4, pp. 571–588, 2019. Artificial Intelligence Applications and Innovations. AIAI
[47] M. A. Khan, I. Ashraf, M. Alhaisoni et al., “Multimodal brain 2020. IFIP Advances in Information and Communication
tumor classification using deep learning and robust feature Technology, vol 583Springer, Cham.
selection: a machine learning application for radiologists,” [65] I. A. E. Kader, G. Xu, Z. Shuai et al., “Brain tumor detection
Diagnostics, vol. 10, no. 8, p. 565, 2020. and classification on MR images by a deep wavelet auto-
[48] M. Siar and M. Teshnehlab, “Brain tumor detection using deep encoder model,” Diagnostics, vol. 11, no. 9, pp. 1–19, 2021.
neural network and machine learning algorithm,” in 2019 9th [66] E. Aarthi, S. Jana, W. G. Theresa et al., “Detection and classifi-
International Conference on Computer and Knowledge Engi- cation of MRI brain tumors using S3-DRLSTM based deep
neering (ICCKE), pp. 363–368, Mashhad, Iran, 2019. learning model,” International Journal of Electrical and Elec-
[49] M. Li, L. Kuang, S. Xu, and Z. Sha, “Brain tumor detection tronics Research, vol. 10, no. 3, pp. 597–603, 2022.
based on multimodal information fusion and convolutional [67] N. A. Samee, N. F. Mahmoud, G. Atteia et al., “Classification
neural network,” IEEE Access, vol. 7, pp. 180134–180146, framework for medical diagnosis of brain tumor with an effec-
2019. tive hybrid transfer learning model,” Diagnostis, vol. 12, no. 10,
[50] MatLab, Version 9.2 (R2017a), The MathWorks Inc., Natick, p. 2541, 2022.
Massachusetts, 2017.
[51] E. Frank, M. A. Hall, and I. H. Witten, The WEKA workbench.
Online appendix for “Data mining: practical machine learning
tools and techniques”, Morgan Kaufmann, Fourth Edition edi-
tion, 2016.
[52] https://www.kaggle.com/datasets/sartajbhuvaji/brain-tumor-
classification-mri.
[53] https://www.kaggle.com/datasets/ahmedhamada0/brain-
tumor-detection?-Resource=download.
[54] M. Coşkun, Ö. Yildirim, U. Ç. A. R. Ayşegül, and Y. Demir,
“An overview of popular deep learning methods,” European
Journal of Technique, vol. 7, no. 2, pp. 165–176, 2017.
[55] https://neurohive.io/en/popular-networks/alexnet-imagenet-
classification-with-deep-convolutional-neural-networks/.
[56] K. Simonyan and A. Zisserman, “Very deep convolutional net-
works for large- scale image recognition,” 2014, http://arxiv
.org/abs/1409.1556.
[57] X. Han, Y. Zhong, L. Cao, and L. Zhang, “Pre-trained Alexnet
architecture with pyramid pooling and supervision for high
spatial resolution remote sensing image scene classification,”
Remote Sensing, vol. 9, no. 8, p. 848, 2017.
[58] A. M. Carvalho, T. T. Roos, A. L. Oliveira, and P. Myllymäki,
“Discriminative learning of Bayesian networks via factorized
conditional log-likelihood,” Journal of Machine Learning
Research, vol. 12, pp. 2181–2210, 2011.
[59] N. Friedman, D. Geiger, and M. Goldszmidt, “Bayesian net-
work classifiers,” Machine Learning, vol. 29, no. 2/3, pp. 131–
163, 1997.
[60] S. P. Rajamohana, K. Umamaheswari, and R. Karthiga, “Senti-
ment classification based on LDA using SMO classifier,” Inter-