INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIII, Issue IX, September 2024

www.ijltemas.in Page 77

Detection of Brain Tumor using Medical Images: A Comparative

Study of Machine Learning Algorithms – A Systematic Literature

Review

Solomon Antwi Buabeng

1,2

4

University Clinic, University of Energy and Natural Resources Sunyani, Ghana

DOI: https://doi.org/10.51583/IJLTEMAS.2024.130907

Abstract: Background: Brain tumors are a significant global health concern impacting both adults and children. Tumors are

characterized by abnormal or excessive growth resulting from uncontrolled cell division. Diagnosing brain tumors poses various

challenges, including limited funding, a shortage of qualified professionals, and insufficient access to medical facilities in remote

regions. Different learning techniques for detecting brain tumors have been developed due to their ease of use, cost-effectiveness,

and non-invasive nature, in contrast to other invasive methods.

Methods: This research conducts a systematic literature review to explore modern trends and concepts of machine learning in

imaging, and clinical evaluations in medical fields related to brain tumor detection.

Keywords: Classification, Artificial Intelligence, Deep Learning, Machine Learning.

I. Introduction

The International Agency for Research on Cancer (IARC) reports that approximately 126,000 people are diagnosed with brain

tumors annually, with an estimated 97,000 deaths (Al-Tamimi & Sulong, 2015). Abnormal human nervous system growth poses

significant health risks (Sarkar, Alahe, & Ahmad, 2023). Brain tumors are among the most severe health challenges today due to the

development of uncontrolled destructive cells in the nervous system. Early detection and treatment are crucial, as brain tumors can be

particularly dangerous if not addressed promptly (Krishnapriya & Karuna, 2017). The World Health Organization (WHO) identifies

malignant brain tumors as destructive and fatal neoplasms with high mortality rates across all age groups (Xie et al., 2022).

Annually, 9.6 million people worldwide die from brain cancer, according to the WHO (Vimala, Srinivasan, Mathivanan, Jayagopal,

brain tumors are used for diagnosis and treatment planning (Dong et al., 2011). Causes include inheritance, radiation exposure,

metastasis, and medical history, with symptoms such as headaches, nausea, and other standard symptoms of brain tumors and

cancers (Raghavapudi, Singroul, & Kohila, 2021). Symptoms also include seizures, vision changes, headaches, hearing issues,

balance problems, cognitive difficulties, and sudden mood changes like aggression and hallucinations (Elshaikh et al., 2021).

Common detection techniques include MRI, X-ray, and CT scans, which are critical for early brain tumor detection and effective

treatment (Biswas & Islam, 2023). Non-invasive techniques are vital in treatment processes (Rasool et al., 2022). Manual MRI

diagnosis can be time-consuming and prone to errors. Deep neural frameworks automate complex medical processes, aiding

healthcare professionals in diagnosis (Kuraparthi, Reddy, Sujatha, Valiveti, & Duggineni, 2021). Non-invasive, smartphone-based

devices show promise for addressing healthcare concerns related to brain tumors (Jia, Shkolyar, & Laurie, 2021). Various studies

have developed robust, non-invasive techniques for brain tumor detection and diagnosis, utilizing medical imaging and machine

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIII, Issue IX, September 2024

www.ijltemas.in Page 78

Table1 Research questions underlying this systematic literature review

Positron Emission Tomography (PET), and biopsies (tissue sample analysis). However, these methods, particularly MRI, can be

time-consuming and heavily reliant on the expertise of the operator (Mangla, 2022; Shohag, Aktar, Science, & Imtiaz, 2015; Ali et

al., 2022; Amin, Sharif, Haldorai, Yasmin, & Sundar, 2022). MRI is the most commonly used tool for detecting and diagnosing brain

tumors, yet the interpretation of MRI images can be challenging due to the complexity and variability of brain anatomy (Suchetha,

Bhat, Hegde, Mallikarjun, & Karthik, 2023). In treating brain tumors, neurosurgical resection is often the first step (Vermeulen et al.,

2023).

Deep learning, particularly Convolutional Neural Networks (CNNs), has shown superior performance in computer vision tasks,

including medical image analysis (Miah et al., 2024). These deep-learning models continuously mine significant image features,

evaluate patterns, and classify these features (Ali et al., 2022). Computer-aided diagnostics (CAD) has significantly progressed with

machine learning and deep learning techniques in medical image detection (Islam et al., 2022; Anagun, 2023). Since the introduction

This systematic literature review makes the following contributions:

Examined 20 original papers on machine-learning techniques for brain tumour detection using medical images, providing

comprehensive knowledge in this domain.

Offers a detailed analysis of medical image segmentation, including:

(a) Evaluation techniques

(b) Machine learning techniques

(c) Dataset sizes used in research

(d) Performance of the adopted ML models

Provides findings that will guide future research and development in this field.

II. Method

This narrative review critically examines and analyses the current information on the application of machine-learning techniques for

brain tumour detection. The review adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA)

standards. The PRISMA checklist was followed, and the Mendeley data repository was registered on December 12, 2019. Research

articles were sourced from Research Gate, Google Scholar, PubMed, and Directory of Open Access Journals databases, covering the

Only English-language literature was used. Papers lacking sufficient data for outcome assessment or those not including medical

images were excluded.

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIII, Issue IX, September 2024

www.ijltemas.in Page 79

Artificial intelligence and deep learning are increasingly being integrated into various technological domains. In medical imaging,

there has been a significant uptick in the use of AI, particularly convolutional neural networks (CNNs), which mimic the neurons in

the human visual cortex (Gordon, 2021). CNNs can process vast amounts of data more quickly than humans, and in the context of

brain tumors, they have demonstrated performance comparable to expert levels (Gordon, 2021). Recently, CNNs have become more

prevalent in brain tumor classification due to their outstanding performance and extremely high accuracy in research environments

(Xie et al., 2022).

Additionally, it outlines the screening of abstracts and titles, as well as the inclusion and exclusion of full papers following the search

across different databases.

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIII, Issue IX, September 2024

www.ijltemas.in Page 80

Table 2: This table presents the performance metrics of various machine learning techniques used for brain tumor detection, along

with detailed information about each model. Furthermore, it provides an overview of current studies focusing on machine learning

models and their respective summaries.

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIII, Issue IX, September 2024

www.ijltemas.in Page 81

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIII, Issue IX, September 2024

www.ijltemas.in Page 82

frequently employed in medical imaging problems, are predominantly used in the detection of brain tumors and medical image

analysis due to their high accuracy performance in research contexts.

recognition and ability to discern complex data relationships.

Convolutional Neural Networks (CNNs) represent a class of deep learning methodologies predominantly utilized for data analysis

and visualization, requiring minimal preprocessing strategies. CNNs have found extensive application across various domains,

including natural language processing, image classification, and medical imaging. Presently, machine learning, particularly deep

learning techniques, is adept at discerning patterns within medical images, as it excels in extracting and synthesizing knowledge from

data rather than relying solely on scientific texts. Machine learning techniques have evolved into potent tools for medical

applications, playing a crucial role in enhancing performance across diverse fields such as tissue segmentation, diagnosis, and

molecular identification (Amin et al., 2022; Saeedi et al., 2023).

In a study conducted by Saad, Suliman, Bitar, and Bshara (2023), a hybrid model was employed for brain tumor detection using MRI

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIII, Issue IX, September 2024

www.ijltemas.in Page 83

image feature extraction (Aafreen et al., 2022). In this context, Hossain, Shishir, Ashraf, and Alpha (2019) proposed a CNN

architecture for brain tumor detection consisting of 5 layers, 32 convolutional filters of 3*3, and 3 channels, achieving an accuracy of

97.87%. Kuraparthi et al. (2021) introduced a deep neural network for brain tumor detection, employing CNN models such as Alex

Net, ResNet50, and VGG16 on Kaggle and Brats publicly available datasets. The proposed technique achieved accuracies ranging

from 98.28% to 99.0% with and without data augmentation, respectively, with the ResNet model. Miah et al. (2024) investigated

brain tumor detection using clustering and softmax for MRI image analysis. The performance of CNN was compared with Decision

validation (Menze et al., 2015; Deepak & Ameer, 2019) can enhance their effectiveness.

Data augmentation is a commonly employed technique to enhance the generalization of deep learning models. Smaller dataset sizes

often lead to overfitting, where the model memorizes the training data (Nalepa, Marcinkiewicz, & Kawulok, 2019). (Biswas & Islam,

2023) conducted research using the CNN-SVM technique for brain tumor detection with a large dataset but without applying data

augmentation, while (Gutta et al., 2024) utilized a sizable dataset obtained from Region of Interest. Conversely, (Swarup et al., 2023)

employed a large dataset but utilized only two convolutional neural networks for brain tumor detection. Additionally, (Shohag et al.,

2015), (Kibriya et al., 2022), and (Miah et al., 2024) utilized much larger datasets for detection but did not specify the origin or

source of the datasets.

However, (Almadhoun & Naser, 2022), (Sarkar et al., 2023), and (Gutta et al., 2024) addressed these dataset gaps, as outlined in

section 1.1. The study by (Saeedi et al., 2023) provided comprehensive details on data collection, including the source, preprocessing

medical images for brain tumor detection. However, it neglects materials not published in reputable peer-reviewed scientific journals,

as these are not indexed and accessible through the databases searched.

III. Conclusion And Future Works

The utilisation of machine learning in medical image detection, such as for brain tumors, is cost-effective and highly efficient.

Timely and accurate decision-making is crucial for medical practitioners, and machine learning excels in this domain. A study

conducted by Naveen et al. (2021) achieved an accuracy of 95.42% in brain tumor detection using a convolutional neural network

(CNN), while Abdelgawad (2023) reported a slightly lower accuracy of 93.3% with CNN. Tasci et al. (2022) employed a soft-

voting-based ensemble technique on TCGA and CGGA datasets, yielding accuracies of 87.606% and 79.668%, respectively.

However, Abbood et al. (2021) outperformed Tasci et al. (2022) with an accuracy of 95.58%.

Furthermore, Samee et al. (2022) achieved a 93.29% accuracy with a neural network, slightly lower than Anaraki et al.'s (2018) CNN

Data augmentation is crucial in medical imaging to mitigate biases from uneven class distribution. It enhances the performance of

machine learning models, facilitating efficient tumor detection and aiding in effective medical treatments. However, a significant

limitation lies in using small datasets, which perform better and prevent overfitting when augmented with larger datasets.This study

aims to illustrate the effectiveness of machine learning models in disease detection through a comprehensive analysis of medical

images. It underscores the importance of utilizing machine learning to reduce costs and enable informed medical decisions.

The systematic literature review evaluated various machine learning models for brain tumor detection, analyzing dataset sources and

augmentation techniques. Alex Net emerged as the most frequently used model, followed by ResNet and VGG16. The study

recommends augmenting medical image datasets to improve accuracy and mitigate overfitting.

Referencing

1. (Samee et al., 2022. (2022). Building Efficient Neural Networks For Brain Tumor Detection. 6(11), 222–235.

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XIII, Issue IX, September 2024

www.ijltemas.in Page 84

3. Aafreen, S., Zarreen, I., Ahemad, A., & Razzaque, P. A. (2022). Brain Tumor Detection using Deep Learning. 4(4), 41–45.

4. Abbood, A. A., Shallal, Q. M., & Fadhel, M. A. (2021). Automated brain tumor classification using various deep learning

5. Abdelgawad, M. I. M. M. M. and A. (2023). A Deep Analysis of Brain Tumor Detection from MR Images Using Deep