IAES International Journal of Artificial Intelligence (IJ-AI)

Vol. 12, No. 1, March 2023, pp. 469~477

ISSN: 2252-8938, DOI: 10.11591/ijai.v12.i1.pp469-477  469

Comparative evaluation for detection of brain tumor using

machine learning algorithms

Shahab Wahhab Kareem1,2, Bikhtiyar Friyad Abdulrahman1, Roojwan Sc. Hawezi1, Farah Sami
Khoshaba1, Shavan Askar1, Karwan Muhammed Muheden1, Ibrahim Shamal Abdulkhaleq1
1
Department of Information System Engineering, Erbil Technical Engineering College, Erbil Polytechnic University, Erbil, Iraq
2
Department of Information Technology, College of Engineering and Computer Science, Lebanese French University, Erbil, Iraq

Article Info ABSTRACT

Article history: Automated flaw identification has become more important in medical
imaging. For patient preparation, unaided prediction of tumor (brain)
Received Sep 20, 2021 detection in the magnetic resonance imaging process (MRI) is critical.
Revised Aug 16, 2022 Traditional ways of recognizing z are intended to make radiologists' jobs
Accepted Sep 14, 2022 easier. The size and variety of molecular structures in brain tumors is one of
the issues with MRI brain tumor diagnosis. Deep learning (DL) techniques
(artificial neural network (ANN), naive Bayes (NB), multi-layer perceptron
Keywords: (MLP)) are used in this article to detect brain cancers in MRI data. The
preprocessing techniques are used to eliminate textural features from the
Brain tumor brain MRI images. These characteristics are then utilized to train a machine-
Brain tumor detection learning system.
Image acquisition
Machine learning algorithm
Magnetic resonance imaging This is an open access article under the CC BY-SA license.

Corresponding Author:
Shahab Wahab Kareem
Department of Information System Engineering, Erbil Technical Engineering College, Erbil Polytechnic
University
Kirkuk Road, 44001, Erbil, Kiurdistan, Iraq
Email: [Link]@[Link]

1. INTRODUCTION
In the age of e-healthcare and information technologies, physicians will deliver quality health care
to their patients [1]. This research discusses the problems relating to the segmentation and management of
dysfunctional usual tissues [2], [3], and brain tissues, such as grey matter (GM) processes, detection of white
matter (WM), and controlling of cerebrospinal fluid (CSF) from applicable magnetic resonance (MR)
imaging techniques and photos utilizing a support vector machine (SVM) classifier and applicable feature
extraction technique [4]. The tumor is defined as an unregulated and uncontrolled growth of cancer cells or
expansion grid [5]. Concept of brain tumor is considered an uncontrolled or unmaintained growth of cancer
or malignant cells. In simple forms, a brain tumor can be of two types, benign growth and forms of malignant
growth [6]. For patients, forms of benign brain tumors are structurally standardized including no active
(cancer) cells [6]. In patients, malignant brain tumors are structurally non-uniform (heterogeneous) including
forms of active cancer cells. Gliomas and meningioma are manifestations of forms of low-grade tumors and
growth known as benign tumors [7].
On the other hand, glioblastoma and astrocytoma are considered high-grade growth tumors
classified as malignant tumors [8]. According to World Health Organization (WHO) and, brain tumor
association of American Brain Tumor Speculation (ABTA), the most popular and accepted mode of
classification scheme utilizes grade I and grade IV scales to distinguish different forms of benign growth and
malignant tumor growth forms [9]. Benign cancers fell below stages of grade I (categorized) and grade II

Journal homepage: [Link]

470  ISSN: 2252-8938

glioma growth on this scale, and different forms of malignant tumors growth fall below grade categorization
of III growth and graded under IV glioma growth.
Grade I and II tumors tend to grow slowly, while grade II tumors tend to grow rapidly [10]. If a low-
grade brain tumor is not handled, it will progress to a high-grade brain tumor, which will then become a
malignant form of brain tumor showing irregular and uncontrolled growth. Patients with different forms of
grade II gliomas can be watched closely and have the magnetic resonance imaging (MRI) or computer
tomography (CT scans) every 6 to 12 months [11]. Every person, at any age, may be affected by a brain
tumor, and the effects on the body vary from person to person. Low-grade categorized as I and II glioma
(uncontrolled growth) benign tumors are considered to be curable following a complete surgical procedure in
this process, while grade III and IV forms of malignant brain tumors may be treated with different forms of
radiotherapy procedure, chemotherapy treatment, or a combination of the two treatment procedures [12].
Both type III and IV glioma, also known as anaplastic astrocytoma, are used in the term malignant
glioma. Anaplastic astrocytoma is a mid-grade tumor with irregular or uncommon growth features and a
higher growth index than most low-grade tumors [13]. Glioblastoma is also the most dangerous form and
type of astrocytoma seen among the patients, as well is categorized as the highest grade glioma in this
category. Glioblastoma is distinguished from other tumor types by the abnormal and uncontrolled forms of
accelerated growth of different types of blood vessels and the presence of uncontrolled types of necrosis
(dead cells) around the tumor segment in the patient [14]. Glioblastoma which is an extremely uncontrolled
form of growth, at the grade IV tumor stage, is a fast-growing and highly contagious, and malignant tumor
type as contrasted to other tumor groups via different types of medical imaging methods, segmentation is
used to detect infected tumor tissues [15]. Segmentation is a way of separating and differentiating a picture
into different blocks in terms of sharing common properties, such as color, shape, growth, structure, contrast,
brightness, development, border, and grey level in the patients [16]. It is a significant step in the processing
of pictures. The method of separating and differentiating different forms of tumor tissues such as oedema and
dead cells formation, from normal brain tissues and healthy tumors is categorized as different developmental
stages for WM, GM, and CSF using MR images or other imaging techniques is known as segmentation of
brain tumors [17].

2. THE COMPREHENSIVE THEORETICAL BASIS

2.1. Detection and identification of brain tumors in MRI imaging
The MRI brain images to detect brain tumors are reprocessed using a median filter, during which the
brain tissue is analyzed through morphological segmentation and tumor extraction techniques [15]. This is a
great way to provide an accurate MRI picture of the tumor, (detection, determination, identification, and
classification of brain tumors in MR images by) explains how to use the neuro-fuzzy classifier to extract the
tumor portion of the brain image [18], after which features were computed from the extracted texture using
the gray level development and formation of co-occurrence procedure matrix process (GLCM) neural
networks and segmentation systems detected tumors [19]. The patient controlled analgesia (PCA) is used for
function extraction and the derived features of MRI. And the typical recognition rate is around 88.2%,
whereas the best recognition rate is about 96.7% In their attempt to use picture the brain tumors, Rajesh and
others suggested a segmentation strategy, which employs the probiotic neural network (PNN) to classify the
MRI scans [19]–[21]. The PNN recommended handling the naming of brain tumors with greater finesse
Suchita and Lalit, developed an unsupervised learning procedure for photographs of the human brain through
the use of the positron emission tomography (PET) scanner [22]. Until they are analyzed, the MRI images are
thresholded, noise is removed, and the tumor is segmented [23]. The brain's structure is detected and
maintained by using and incorporating the GLCM method which is widely used for image processing,
development, determination, and subsequent self-organizing mapping (SOM).

2.2. Brain tumor types

A tumor is generated by a rise in the number of cells without being balanced out by a decrease in the
number of neuron deaths a tumor in the human brain can exert a strong influence over the whole body,
including the skull, and can lead to impairment in the human health. The numbers of brain tumors increase
daily [21]. To improve treatment, detection early on is vital for all brain tumors, as it is for all diseases. Early
detection of brain tumors is also discovered through the use of MRI. MRI does not include the introduction
of drugs, is painless, is done with low doses of radiation, and does not expose the patient. The normal MRI
scanning sequence consists of axial, coronal, and sagittal views. The MRI imaging allows for three imaging
modalities to be used to provide more detailed details on the tumor structure, tissue, and thickness [24]. The
three-dimensional (3D) phase MRI scan can be run in three separate ways. A 3D brain MRI scan of
biological tissue with a single-sequence MRI may not be able to distinguish between these two kinds of

Int J Artif Intell, Vol. 12, No. 1, March 2023: 469-477

Int J Artif Intell ISSN: 2252-8938  471

tumors. MR imaging has two sequences, T1- and T2 weighted. T1-weighted MRI helps identify brain tumors
that have shifted to the cerebral spinal fluid spaces. The contrast used in T2-weighted MR imaging is used to
depict pathology on the other side is heavy. Brain tumors may be found nearby. Thus, the tumors are given
names based on which will classify them according to the types of cells. Tumors are usually found in the
brain and pituitary gland as the name indicates, glioma is a disease that affects the brain and the spinal cord.
Gliomas originate in the glial help cells of the brain. It's a tumor that occupies the brain, but is still able to
shape itself in the membrane of the dura, and arises from the membrane. One of the most often occurring
brain tumors in adults is of this kind is found to include between 15-20% of brain tumors [21]. Finally, a
pituitary tumor grows in the pituitary gland, which establishes hormonal equilibrium, which in turn regulates
the activity of other glands, which includes a portion of the menstrual cycle. T1-weighted MRI with three
differentiating the tumor groups using MRI to diagnose brain tumors leads to wasted time and misdiagnoses
[25]. Deep learning (DL) and image recognition algorithms are already used in recent computerized MR
diagnostic programs to cut both the time and the margin of error a deep learning-based, object recognition
technique utilizes multilayered neural networks and is part of the kind of machine learning called artificial
intelligence. It is possible for DL to learn from the same kind of data as machine learning algorithms do:
images, video, and text as well as when used by conventional techniques [24], but these are significantly
different approaches to solving the problem. In this research, we seek to build a web-based program that is
capable of using an in-depth T1 MRI scan for glioma, meningioma, and pituitary adnexa classifications
through DL so are suspected that health practitioners and scientists can now easily diagnose brain tumors
because of the free web-based tools created Often, this web-program can be used as a diagnostic tool for
brain tumor classifications (i.e., glioma, meningioma, pituitary). The experimental findings show that all
measured indicators can distinguish the various categories of brain tumors on the training dataset, which is
significantly higher than 98%. However, for all indicators the score is greater than or equal to 91%, except
for sensitivities and the major complication or comorbidity (MCC) for meningioma. On the preparation and
evaluation points, the convolutional neural network (CNN-based) model can accurately identify the various
brain tumor forms [26], [27]. An intriguing thesis on CNN is researching brain tumor classifications on
public MRI pictures, which used the technique of DL to a group over 3,400 different T1-weighted images of
the same subjects into 233 and 3,000 exclusive models. When run on two different datasets, the evolved
method produces overall accuracy of 96.13% and 98.7%.

2.3. Method for detection

Results from an empirical research survey show that automatic brain tumor diagnosis is critical
where there is a human existence at stake [21], [28]. Figure 1 shows the usable method for detecting brain
tumors, starting with image acquisition, preprocessing, extraction and segmentation. Function extraction and
classification is the technique that is widely used to diagnose cancer in MRI photos using a machine learning
algorithm.

Figure 1. Usable method for brain tumor detection in MRI imaging

2.3.1. Image acquisition stage for MRI

In the initial stage, the MRI images for brain scanning are gathered, which also works as the input
for the pre-processing stage for the overall interconnected system [22]. In this stage, different forms of
sample images are collected for relevance. In the image acquisition and development stage, consequent
management of the performance metrics is ensured [29]. In Figure 2 shows the normal brain MR image refers
to non-tumour images.

2.3.2. Preprocessing
Preprocessing offers gain by refining some of the image functionality for eventual processing [24].
Until processing the MRI pictures, the following measures are applied: As seen in Figure 3(a), the red, green,
blue (RGB)-formed MRI image is first normalized into greyscale and then subjected to median filtering
Figure 3(b). Since more precise analysis is needed, the noise has been filtered out. Canny edge identification

Comparative evaluation for detection of brain tumor using … (Shahab Wahhab Kareem)
472  ISSN: 2252-8938

and detection method as depicted in Figure 3(c). To segment the picture, the edge detected image is needed.
The picture in Figure 3 is a procedure used to find the tumor's position in the brain's watershed segmentation
Figure 3(d). The segmentation of a picture is a series of segments. Segmentation seeks to reorganize picture
representation [24]. A marked picture in the product of watershed division in graphic, all the objects and
tools will have different values will be on the same scale, all the first object's pixels will have value 1, all the
second object's pixels' first, all other objects' pixels will have a second value [30], and so on This chart shows
in Figure 3 illustrates numerous preprocessing techniques applied to brain MR images.

Figure 2. Brain MR image (samples)

(a) (b) (c) (d)

Figure 3. Stages in preprocessing of MRI images (a) original image, (b) filtered image, (c) edge detected
image, and (d) segmented image

2.3.3. Feature extraction

When an algorithm handles massive volumes of input data and dramatically simplifies it down to
only a few functions, it is claimed to reduce to a function vector [31]. Feature extraction involves the
transformation of input data into a series of functions of function extraction the texture segmentation
approach is used, and the extracted forms of textural features are depicted in the segmented MR image [32].
The GLCM process is a versatile tool of exceptional efficiency, used to remove these characteristics [33].
The GLCM tools for the identification of texture extraction process is very capable and used for the
development and identification process as it uses fewer grey levels while improving the overall classification
[34]. To further identify and distinguish between usual and abnormal conduct, the gray level measurement
and co-occurrence (GLC) features are used [35]. There are details to be found in texture about the way
objects are placed on the floor. Grey tone spatial features have more general applicability to picture
classification relevance.

2.3.4. Classification
In certain cases, multi-layer perceptron (MLP) and a naive Bayes (NB) algorithm are used to
perform pattern recognition on photographs of the brain [36] to help them learn and make better decisions
This method involves creating an artificial neural network (ANN) [37] in which multiple layers map inputs to
a single output variable that feeds to a second one or multiple subsequent layers until a suitable response is
reached. Each neuron in the MLP is regulated by its special activation mechanism [29]. A feedback circuit is
known as a feedforward since it has no feedback [36]. It's all in the journal. When the goals for success and
actual results align, learning occurs by changing the association weights [38]. Because of this negative
feedback, [39] the system is known as derisively as 'backward pathogenesis'. As a goal, the total weight
values must be minimized such that each edge has a lower chance of error. Figure 4 show result of all the
mentioned stages, the result of the preprocessed image Figure 4(a), k-means segmented image result shown
in Figure 4(b), fuzzy c-means segmented image shown in Figure 4(c), clear border result in Figure 4(d), and
feature extracted image in Figure 4(e).

Int J Artif Intell, Vol. 12, No. 1, March 2023: 469-477

Int J Artif Intell ISSN: 2252-8938  473

(a) (b) (c) (d) (e)

Figure 4. Classification for different MRI outputs, (a) original image, (b) k-means segmented image,
(c) fuzzy c-means segmented image, (d) clear border, and (e) feature extracted image

3. METHOD
3.1. Methods and eligibility criteria
A systematic review and comparative study have been conducted based on the acceptability and
relevance of the user data and speculation for the research process. Different methodological perspectives of
the research have been adopted effectively to analyze the findings. For allocating research and maintaining
the research process [40], research philosophy, research design, and ethical processes have been selected
aligning with the requirements of the research.

3.2. Search criteria

The search technique in this research is an organized structure of main words used to search the
target database. The search approach incorporates findings and the core principles of the target search query
to obtain correct answers. Prominent databases such as the Institute of Electrical and Electronics Engineers
(IEEE), Journal of Science and Technological Researches (JSTR), and the International Institute for Science,
Technology and Education (IISTE) have been included in the search strategy. The search criteria prioritized
the relevance of the topic with the objectives of the current research. Search criteria are based on: i) articles
published between 2012 and 2021 have been selected for this research; ii) databases such as word elektrik
browser (WEB) of science, IEEE, JSTR, IISTE, and employee database (EMBASE). have been taken into
consideration for this research; and iii) keywords such as brain tumor, screening brain tumor, detection
techniques, and MRI image. have been used for the search purpose, using or, and criteria based on the
assessment criteria and requirements of the research.

3.3. Ethics
Different ethical contexts have also been addressed in this research to assess the findings effectively.
Selection of the articles and journals have been conducted based on the reliability and acceptability of the
published articles. To assess the ethical contexts, different criteria for the research processes have been
addressed effectively.

3.4. Protocols and procedures

In this research, protocols, and procedures have been selected carefully to maintain the acceptability
and reliability of the research [40]–[46]. In this systematic review of articles, different dimensions for
reliability and validity have been addressed accordingly. To analyze the concepts of the evaluative tactics and
strategies to detect brain tumors from screening MRI images, a comparative tool for the selection of articles
has been taken into consideration. Appropriate procedures and protocols have been selected in this regard.

4. RESULTS AND DISCUSSION

4.1. Excerpts from prominent research
To present some of the comparison points, Table 1 shows the details of the comparison of some
research work in the field of the machine learning approach for detecting brian tumours. The analysis is
based on the outcomes of the individual research works. Table 1 shows comparison of several papers based
on the detection mechanism.

4.2. Discussion
Cancer that has developed in the brain is called a brain tumor [31]. Benign and malignant tumors
may be located in the brain. Clustering is commonly used in the diagnosis of diseased tissues. It is recognized
that low-grade I and high-grade gliomas are considered to be curable [32]. Tumors with grade III and IV
properties are usually are well-tolerated by radiotherapy and chemotherapy [33]. Glioblastoma is the most
dangerous and the most category form of astrocytoma [35]. Grade IV glioma is a very malignant tumor with
Comparative evaluation for detection of brain tumor using … (Shahab Wahhab Kareem)
474  ISSN: 2252-8938

a high incidence of grade V metastasis. the procedure of differentiating the tumor from normal brain tissue,
for example.
Imaging or pictorial analysis is central to the modern body of knowledge. It's very hard to get your
hands around the procedure of making some kind of graphic presentation. Unless otherwise stated,
visualizations of the feature are referred to as medical imaging in this section. Algorithms can resolve
complicated medical imaging issues to some degree. An embedded co-processor uses a converter that
converts an algebra to a hard-wired arithmetic algorithm. Because the geometry of an application's artifacts is
quite good, very effective, it would likely be very inefficient through the use of computer graphic modeling.
Since conformal geometry does have a defined geometry at a coordinate-free stage, it might be able to meet
these problems to a significant degree. Potential is greater than the number. a procedure that is effective and
precise in terms of both detection and measurement of three- and registration of 3D shapes you don't have to
be overly complicated with free curves and free surfaces. The iterative closest point (ICP) approach will
accommodate the most freedom and is built on iterative calculations. There are around a million sums or
contractions in the framework profile. A 4-dimensional (4D) image of elements or 4 sets of 4 elements
conformal geometric algebra is used for the initial detection of the tumor when running on a general-purpose
processor. Neural networks can be used to detect some anatomical features of the tumor this is effective when
applied to colored photographs as well. The processing time required for this algorithm can be very short, as
long as it is there are few photographs. Even, as well as the tumor volume, we detected with the aid of the
form and diameter of the detected tumor using the methodology is more efficient, but it cannot overcome
high computational costs.

Table 1. Comparison among the research works

Authors Findings
This paper proposes and incorporates machine learning algorithms to assist in brain tumor detection.
It is possible to increase the accuracy and granularity of the data collection by taking into account all
texture and force characteristics in addition to the collected information. Concepts are established by
the GLCM. In this proposed work, energy is introduced into the skin, along with image contrast,
Sharma et al. [12]
dignity, and homogeneity. Videos for both discrimination and accuracy purposes, the multi-layer
peron grouping, distinction, and machine learning methods have a determinant approach that targets
the top 2.2% of both, and the Bayesian process returns a classification and accuracy rating of 97.6%
(accuracy goal).
The creative wave transformation (BWT) was employed in this study to increase the speed and reduce
the amount of effort required for image segmentation. The surgical or therapeutic segmentation and
diagnosis of tumors using MRI is a big issue, but it is a complicated and lengthy procedure based on
skills and knowledge that radiologists cannot help with. Therefore, computer-assisted technology use
becomes critical to meeting these challenges. SVM-dependent features are isolated from each
segmented tissue to improve classification precision and efficiency. Based on the accuracy,
Bahadure et al. [13] sensitivity, specificity, and basis of dice coefficients, this technique has been evaluated and
extensively studied to analyze its efficacy and reliability for brain imaging studies. "They showed that
the proposed method can discriminate between healthy and unhealthy tissue images, with 96.51%
precision, 94.2% sensitivity, and 97.72% accuracy." Overall, the results showed a dice coefficient of
0.82, with the computer extraction of the tumor yielding more substantial overlap with the manually
derived tumor area. In comparison to existing state-of-of-the-the-the-art approaches, the results
highlight the accuracy and precision parameters.
Medical image recognition, medical image interpretation, and bioinformatics have seen a resurgence
with the advent of DL in recent years. Because of this, DL has had a big influence on how
malignancies can be detected and quantified in various areas, including brain tumors, the liver, the
belly, and the retina. Since it is extremely relevant to brain tumor science, we hope that it can shed
light on major principles of DL (e.g., segmentation, classification, prediction, evaluation). Data layers
Nadeem et al. [14] of computational models are processed using DL often incorporating several levels of abstraction.
This essay provides a summary of some of the important developments in the study field (i.e., DL in
brain tumor analysis). There have been several various forms of areas studied and placed to the test,
as well as a taxonomy that has preserved continuity throughout the literature. The following is a
timely discussion of DL key weaknesses in this modern technology environment offered as well as a
collection of tools on how it can be enhanced for usage in the new media market.
A brain tumor cannot be managed in the latest stages, since it is the most dangerous disease of all. It
should be located by an MRI. What happened to the patient, in other words, means that he/she shifted
to stay alive. The words "brain tumor" and "glial glioma" apply to the multiplication of glial cells. In
the majority of cases, it's the deoxyribonucleic acid (DNA) that is broken. In the human brain, there
are almost ten times as many cells as there are in the whole body. Self-damaged brain cells determine
their condition by splitting and reproducing. The MRI produced grey-scale images, in which the
Gore and Deshpande [15]
nuclei of tumors are colored in darker shades of grey to distinguish between stable and diseased
tissue. Given the complexity and lack of uniformity of MRI scans, grey matter-based tumor diagnosis
is not reliable. In this effort, we developed a machine-learning algorithm to find tumors. We used
CNN and ANN to classify the tumor in the brain. Both un-augmented and augmented databases are
included in CNN. The key purpose of the paper is to compare two model predictions about efficiency,
precision, and sensitivity, and to choose the better one.

Int J Artif Intell, Vol. 12, No. 1, March 2023: 469-477

Int J Artif Intell ISSN: 2252-8938  475

5. CONCLUSION
MRI image techniques used to determine different forms of benign and malignant growths have
been an important aspect of the treatment procedures of the patients. In this way, medical professionals can
determine the required level of strategic aspects to understand the processes and aspects needed for the
treatment processes. To analyze the research outcomes, different aspects of the MRI imaging techniques have
been discussed in this report. A brief comparative analysis of the literature has been also conducted to review
the findings. The conceptual models use several levels of abstraction to deal with the data in DL algorithms
in medical imaging, medical processing, research, and bioinformatics, DL has exploded As several other
fields of healthcare have discovered, DL and neural networks have shifted the balance of diagnosis and
inference, these advances have had widespread effects on various forms of pathology, such as liver,
esophageal, brain tumor, and retinal disease, too As major objectives of this article is to examine the popular
applications of DL as it relates to brain tumor research (e.g., segmentation, classification, prediction,
evaluation). This book (i.e., DL in brain tumor analysis) contains several research contributions to the area.

REFERENCES
[1] M. S. Goswami and M. L. K. P. Bhaiya, “Brain tumour detection using unsupervised learning based neural network,” in
Proceedings - 2013 International Conference on Communication Systems and Network Technologies, CSNT 2013, Apr. 2013, pp.
573–577, doi: 10.1109/CSNT.2013.123.
[2] T. Rajesh and R. S. M. Malar, “Rough set theory and feed forward neural network based brain tumor detection in magnetic
resonance images,” in International Conference on Advanced Nanomaterials & Emerging Engineering Technologies, Jul. 2013,
pp. 240–244, doi: 10.1109/ICANMEET.2013.6609287.
[3] S. Rajeshwari and T. S. Sharmila, “Efficient quality analysis of MRI image using preprocessing techniques,” in 2013 IEEE
Conference on Information and Communication Technologies, ICT 2013, Apr. 2013, pp. 391–396, doi:
10.1109/CICT.2013.6558127.
[4] T. A. Assegie, R. L. Tulasi, and N. K. Kumar, “Breast cancer prediction model with decision tree and adaptive boosting,” IAES
International Journal of Artificial Intelligence (IJ-AI), vol. 10, no. 1, pp. 184–190, Mar. 2021, doi: 10.11591/ijai.v10.i1.pp184-
190.
[5] D. Singh and K. Kaur, “Classification of abnormalities in brain MRI images using GLCM , PCA and SVM,” International
Journal of Engineering and Advanced Technology (IJEAT), vol. 1, no. 6, pp. 243–248, 2012.
[6] P. Gadpayleand and P. S. Mahajani, “Detection and classification of brain tumor in MRI images,” International Journal of
Emerging Trends in Electrical and Electronics (IJETEE), vol. 5, no. 1, 2013.
[7] M. Shasidhar, V. S. Raja, and B. V. Kumar, “MRI brain image segmentation using modified fuzzy c-means clustering algorithm,”
in Proceedings - 2011 International Conference on Communication Systems and Network Technologies, CSNT 2011, Jun. 2011,
pp. 473–478, doi: 10.1109/CSNT.2011.102.
[8] K. Sharma and N. Kaur, “Comparative analysis of various edge detection techniques,” International Journal of Advanced
Research in Computer Science and Software Engineering, vol. 3, no. 12, pp. 617 – 621, 2013.
[9] J. Selvakumar, A. Lakshmi, and T. Arivoli, “Brain tumor segmentation and its area calculation in brain MR images using K-mean
clustering and Fuzzy C-mean algorithm,” in IEEE-International Conference on Advances in Engineering, Science and
Management, ICAESM-2012, 2012, pp. 186–190.
[10] R. J. Ramteke and K. M. Yashawant, “Automatic medical image classification and abnormality detection using k-nearest
neighbour,” International Journal of Advanced Computer Research, vol. 2, no. 4, pp. 190–196, 2012.
[11] X. Xuan and Q. Liao, “Statistical structure analysis in MRI brain tumor segmentation,” in Proceedings of the 4th International
Conference on Image and Graphics, ICIG 2007, Aug. 2007, pp. 421–426, doi: 10.1109/ICIG.2007.181.
[12] K. Sharma, A. Kaur, and S. Gujral, “Brain tumor detection based on machine learning algorithms,” International Journal of
Computer Applications, vol. 103, no. 1, pp. 7–11, Oct. 2014, doi: 10.5120/18036-6883.
[13] 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, pp. 1–12, 2017, doi: 10.1155/2017/9749108.
[14] M. W. Nadeem et al., “Brain tumor analysis empowered with deep learning: A review, taxonomy, and future challenges,” Brain
Sciences, vol. 10, no. 2, pp. 1–33, Feb. 2020, doi: 10.3390/brainsci10020118.
[15] D. V. Gore and V. Deshpande, “Comparative study of various techniques using deep learning for brain tumor detection,” in 2020
International Conference for Emerging Technology (INCET), Jun. 2020, pp. 1–4, doi: 10.1109/INCET49848.2020.9154030.
[16] W. Widhiarso, Y. Yohannes, and C. Prakarsah, “Brain tumor classification using gray level co-occurrence matrix and
convolutional neural network,” IJEIS (Indonesian Journal of Electronics and Instrumentation Systems), vol. 8, no. 2, pp. 179–
190, Oct. 2018, doi: 10.22146/ijeis.34713.
[17] H. Mohsen, E.-S. A. El-Dahshan, E.-S. M. El-Horbaty, and A.-B. M. Salem, “Classification using deep learning neural networks
for brain tumors,” Future Computing and Informatics Journal, vol. 3, no. 1, pp. 68–71, Jun. 2018, doi: 10.1016/[Link].2017.12.001.
[18] K. Kamnitsas et al., “Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation,” Medical
Image Analysis, vol. 36, pp. 61–78, Feb. 2017, doi: 10.1016/[Link].2016.10.004.
[19] S. K. Chandra and M. K. Bajpai, “Effective algorithm for benign brain tumor detection using fractional calculus,” in TENCON
2018 - 2018 IEEE Region 10 Conference, Oct. 2018, pp. 2408–2413, doi: 10.1109/TENCON.2018.8650163.
[20] S. W. Kareem, R. Z. Yousif, and S. M. J. Abdalwahid, “An approach for enhancing data confidentiality in hadoop,” Indonesian
Journal of Electrical Engineering and Computer Science, vol. 20, no. 3, pp. 1547–1555, Dec. 2020, doi:
10.11591/ijeecs.v20.i3.pp1547-1555.
[21] A. Triayudi and I. Fitri, “Comparison of the feature selection algorithm in educational data mining,” TELKOMNIKA
(Telecommunication Computing Electronics and Control), vol. 19, no. 6, pp. 1865–1871, Dec. 2021, doi:
10.12928/telkomnika.v19i6.21594.
[22] J. Seetha and S. S. Raja, “Brain tumor classification using Convolutional Neural Networks,” Biomedical and Pharmacology
Journal, vol. 11, no. 3, pp. 1457–1461, Sep. 2018, doi: 10.13005/bpj/1511.
[23] J. Cheng et al., “Enhanced performance of brain tumor classification via tumor region augmentation and partition,” PLoS ONE,
vol. 10, no. 10, pp. 1–13, Oct. 2015, doi: 10.1371/[Link].0140381.
Comparative evaluation for detection of brain tumor using … (Shahab Wahhab Kareem)
476  ISSN: 2252-8938

[24] G. S. Tandel et al., “A review on a deep learning perspective in brain cancer classification,” Cancers, vol. 11, no. 1, pp. 1–32, Jan.
2019, doi: 10.3390/cancers11010111.
[25] G. Litjens et al., “A survey on deep learning in medical image analysis,” Medical Image Analysis, vol. 42, pp. 60–88, Dec. 2017,
doi: 10.1016/[Link].2017.07.005.
[26] L. Szilagyi, L. Lefkovits, and B. Benyo, “Automatic brain tumor segmentation in multispectral MRI volumes using a fuzzy c-
means cascade algorithm,” in 2015 12th International Conference on Fuzzy Systems and Knowledge Discovery (FSKD), Aug.
2015, pp. 285–291, doi: 10.1109/FSKD.2015.7381955.
[27] M. Soltaninejad et al., “Automated brain tumour detection and segmentation using superpixel-based extremely randomized trees
in FLAIR MRI,” International Journal of Computer Assisted Radiology and Surgery, vol. 12, no. 2, pp. 183–203, Feb. 2017, doi:
10.1007/s11548-016-1483-3.
[28] J. R. Fink, M. Muzi, M. Peck, and K. A. Krohn, “Continuing education: Multi-modality brain tumor imaging – MRI, PET, and
PET/MRI,” Journal of Nuclear Medicine, vol. 56, no. 10, pp. 1554–1561, Oct. 2015, doi: 10.2967/jnumed.113.131516.
[29] G. M. Shivakumarswamy, P. V Akshay, T. A. Chethan, B. H. Prajwal, and S. V Hande, “Brain tumour detection using Image
processing and sending tumour information over GSM,” International Journal of Advanced Research in Computer and
Communication Engineering, vol. 5, no. 5, pp. 179–183, 2016, doi: 10.17148/IJARCCE.2016.5543.
[30] Y. Jia, Y. Zhang, and T. Rabczuk, “A novel dynamic multilevel technique for image registration,” Computers & Mathematics
with Applications, vol. 69, no. 9, pp. 909–925, May 2015, doi: 10.1016/[Link].2015.02.010.
[31] T. Kalaiselvi, P. Nagaraja, and P. Sriramakrishnan, “A simple image processing approach to abnormal slices detection from MRI
tumor volumes,” The International journal of Multimedia & Its Applications, vol. 8, no. 1, pp. 55–64, Feb. 2016, doi:
10.5121/ijma.2016.8105.
[32] R. Kumar and K. J. Mathai, “Brain tumor segmentation by modified k-mean with morphological operations,” International
Journal of Innovative Research in Science, Engineering and Technology, vol. 6, no. 8, pp. 16191–16198, 2017, doi:
10.15680/IJIRSET.2016.0608079.
[33] A. Mukaram, C. Murthy, and M. Z. Kurian, “An automatic brain tumour detection, segmentation and classification using MRI
image,” International Journal of Electronics, Electrical and Computational System, vol. 6, no. 5, pp. 54–65, 2017.
[34] S. Patel and D. Rao, “Brain tumor detection in MRI images with new multiple thresholding,” Journal of Network
Communications and Emerging Technologies (JNCET), vol. 7, no. 6, pp. 34–39, 2017.
[35] A. Pawar et al., “Adaptive FEM-based nonrigid image registration using truncated hierarchical B-splines,” Computers &
Mathematics with Applications, vol. 72, no. 8, pp. 2028–2040, Oct. 2016, doi: 10.1016/[Link].2016.05.020.
[36] R. Sharmila and K. S. Joseph, “Brain tumour detection of MR image using Naïve Beyer classifier and support vector machine,”
nternational Journal of Scientific Research in Computer Science, Engineering and Information Technology, vol. 3, no. 3, pp.
690–695, 2018.
[37] M. Havaei, F. Dutil, C. Pal, H. Larochelle, and P.-M. Jodoin, “A convolutional neural network approach to brain lesion
segmentation,” in Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries, Chsm: Springer International
Publishing, 2016, pp. 195–208.
[38] S. Pereira, A. Pinto, V. Alves, and C. A. Silva, “Deep convolutional neural networks for the segmentation of gliomas in multi-
sequence MRI,” in Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries, Chsm: Springer International
Publishing, 2016, pp. 131–143.
[39] “[39].”
[40] D. Yi, M. Zhou, Z. Chen, and O. Gevaert, “3-D convolutional neural networks for glioblastoma segmentation,” pp. 1–8, Nov.
2016.
[41] O. H. Salman, M. F. A. Rasid, M. I. Saripan, and S. K. Subramaniam, “Multi-sources data fusion framework for remote triage
prioritization in telehealth,” Journal of Medical Systems, vol. 38, no. 9, pp. 1–23, Sep. 2014, doi: 10.1007/s10916-014-0103-4.
[42] S. M. R. K. Al-Jumur, S. W. Kareem, and R. Z. Yousif, “Predicting temperature of Erbil City applying deep learning and neural
network,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 22, no. 2, pp. 944–952, May 2021, doi:
10.11591/ijeecs.v22.i2.pp944-952.
[43] S. Kareem and M. C. Okur, “Bayesian network structure learning using hybrid bee optimization and greedy search,” in 3rd
International Mediterranean Science and Engineering Congress (IMSEC 2018), 2018, pp. 1–7.
[44] S. W. Kareem and M. C. Okur, “Pigeon inspired optimization of bayesian network structure learning and a comparative
evaluation,” Journal of Cognitive Science, vol. 20, no. 4, pp. 535–552, Dec. 2019, doi: 10.17791/jcs.2019.20.4.535.
[45] S. W. Kareem and M. C. Okur, “Bayesian network structure learning based on pigeon inspired optimization,” International
Journal of Advanced Trends in Computer Science and Engineering, vol. 8, no. 1, pp. 131–137, 2019, doi:
10.30534/ijatcse/2019/2281.22019.
[46] S. W. Kareem and M. C. Okur, “Evaluation of bayesian network structure learning using elephant swarm water search algorithm,”
in Handbook of Research on Advancements of Swarm Intelligence Algorithms for Solving Real-World Problems, Hershey: IGI
Global, 2020, pp. 139–159.

BIOGRAPHIES OF AUTHORS

Dr. Shahab Wahhab Kareem Shahab Wahhab Kareem his BSc in Control and
Computer Engineering from the University of Technology Baghdad in 2001, an MSc in
Software Engineering from Salahadeen University in 2009, and a Ph.D. in Yasar University
Izmir, Turkey in 2020. My research interests include machine learning and big data. I’m a
lecturer at the Information System Eng. (ISE) Department (2011-till now). He can be contacted
at email: [Link]@[Link] or shahabwk@[Link].

Int J Artif Intell, Vol. 12, No. 1, March 2023: 469-477

Int J Artif Intell ISSN: 2252-8938  477

Bikhtiyar Friyad Abdulrahman BSc computer science Erbil/salahaden university

2008. i am teacher in polytechnic university Erbil. From 2009 to 2013 work as IT in Technical
Enineering College. MSc 2015 BAMU university India. From 2015-till now work as IT
professional in Erbil technical Engineering College. He can be contacted at email:
[Link]@[Link].

Dr. Roojwan Scddeek Hawezi I got a BSc degree from Baghdad University in
Physical science, then I got a higher Diploma in Computer science. Back in 2003, I got my
MSc degree then finally the Ph.D. in 2007 in Computer science from Baghdad too. I'm a
specialist in digital image processing. I'm the former head of the Information System Eng. (ISE)
Department. She can be contacted at email: [Link]@[Link].

Farah Sami Khoshaba She got the [Link]. degree from University of
Technology/Baghdad in the science of Computer and Control engineering in 1999, and worked
as an engineer at the same department, as she was in charge of control laboratory and process
laboratory. Then in 2006, she got the [Link]. degree at the same Department in the
Specialization of Mechatronics. She's a member of academic staff with a grade of lecturer at the
Information System Engineering (ISE) Department (2011-till now). She can be contacted at
email: [Link]@[Link].

Dr. Shavan Askar (Associate Professor of Computer Networks) is currently

faculty member at Erbil Polytechnic University/College of Engineering. He received his PhD
degree in Electronic Systems Engineering from the University of Essex/UK in 2012. He
obtained his MSc (2003) and BSc (2001, Ranked 1st on the college) degrees from the Control
and Systems Engineering Dept. Baghdad. Dr. Askar works in the field of networks that
includes internet of things, software defined networks, optical networks, and 5G. He can be
contacted at email: [Link]@[Link].

Karwan Muhammed Muheden Assistant lecturer started his academic career in

MAY, 2009. From 2009 to 2017 Assistant Lecturer (‫ )معید‬at Information System Engineering
Department/Erbil Technical Engineering College/Erbil Polytechnic University/Erbil-Iraq. MSc.
In Computer Engineering, Wireless Sensor Network (Internet of Things), Firat University,
Elazig–Turkey,2017. Asst. Lecturer in 2020. Erbil Polytechnic University, Erbil Technical
Engineering college, Information System Engineering Dept. He can be contacted at email:
[Link]@[Link].

Ibrahim Shamal Abdulkhaleq BSc in Information System Engineering from

Erbil polytechnic university in 2015. From 2015 to 2022 Engineer in Information System
Engineering. Now MSc student in Information System Engineering from Erbil polytechnic
university. I’m a team leader at Tech Eye Company. He can be contacted at email:
[Link]@[Link].

Comparative evaluation for detection of brain tumor using … (Shahab Wahhab Kareem)