NeuroScan:AI-based Brain Tumor

Detection and Classification for

Improved Diagnosis and Treatment

Abstract— Brain tumors are a devastating and life-threatening

While traditional diagnostic tools such as Computed
ailment, necessitating precise and prompt identification. Tomography (CT) and Ultrasound have been employed,
Magnetic Resonance Imaging (MRI) has emerged as an Medical Resonance Imaging (MRI) has emerged as a
indispensable tool for early detection and assessment of brain more promising avenue. Among radiology methods, MRI
tumors. In recent times, machine learning (ML) has played a stands out by providing clearer and more detailed images
crucial part in enhancing the precision of brain cancer
identification from MRI scans. This document provides a of the brain. It accomplishes this by exploiting the
thorough examination of numerous research papers that center variation in the number of hydrogen atoms within the
on the fusion of MRI and machine learning for brain cancer body. In MRI scans, areas with higher hydrogen atom
recognition. Our central objective is to evaluate the progress density appear brighter. This distinctive feature of MRI
and pinpoint deficiencies within these studies to ascertain the imaging, rooted in the distribution of hydrogen atoms,
most efficient ML algorithm for constructing an optimal model
for brain cancer recognition. Through a systematic scrutiny of can effectively reveal brain tumors. Notably,
the chosen research papers, we assess the techniques, data sets, cerebrospinal fluid (CSF) is the most hydrogen-rich
and consequences of various ML methodologies, while taking region of the brain, appearing as a bright area. Tumors,
into account their merits and limitations. By comparing and due to their elevated hydrogen atom count, also manifest
contrasting the techniques and outcomes of the examined
as bright areas on MRI scans. However, in instances
papers, our goal is to provide well-informed guidance on the
selection of an ML algorithm for brain cancer recognition. Our where brain tumors may not be readily discernible in
review aims to contribute to the ongoing initiatives to enhance conventional MRI images, the utilization of contrast
early diagnosis and management of brain cancer, underscoring agents, such as gadolinium, becomes indispensable.
the pivotal role of machine learning in enhancing medical Gadolinium, a contrast agent commonly employed in
imaging procedures.
MRI, boasts a high glucose content. Given that tumors
Keywords— Magnetic Resonance Imaging; machine avidly absorb glucose, the application of gadolinium
learning; convolutional neural network; diagnostic; datasets; accentuates their presence in MRI images. This
Deep learning; enhancement proves invaluable in rendering previously
inconspicuous tumors more visible.
I. INTRODUCTION
In light of the above considerations, this paper embarks
The mind, often considered the most crucial organ in the on an innovative journey towards the accurate diagnosis and
human physique, operates as the central command center, classification of brain tumors utilizing MRI images and
orchestrating the operations of the body's diverse systems. several machine learning and deep learning methods. We seek
Nevertheless, when the fragile equilibrium of cellular to evaluate and compare the performance of various
division is disrupted, it can lead to the emergence of methodologies, including Convolutional Neural Networks
neoplasms, anomalous growths characterized by (CNNs), Deep Learning, VGG-16, VGG-19, Support Vector
uncontrolled cell proliferation. These growths are categorized Machines (SVM), K-means, and Region-based
into two distinct levels: mild and severe. Mild-grade growths, Convolutional Neural Networks (RCNN), to assess their
typically non-malignant, are non-cancerous and present a suitability for the task at hand.
comparatively lower threat to the individual. In sharp
contrast, severe-grade growths are cancerous, capable of
infiltrating neighboring tissues, dispersing to distant body The subsequent sections of this paper will delve into the
parts, and, if not addressed, even causing mortality [1]. existing research in brain tumor diagnosis, segmentation, and
classification, the methodology employed for tumor
diagnosis and classification, including a detailed exploration
Regretfully, cerebral neoplasms have caused considerable of the CNN and deep learning methods, as well as the
distress, particularly among children, with the year 2016 utilization of VGG-16, VGG-19, SVM, K-means, and
witnessing cerebral neoplasms as the primary cause of death RCNN. We will also discuss the dataset and its division, the
in this age group. Furthermore, cerebral neoplasms also stand presentation and comparison of results, and a conclusion that
as the third most frequent cancer diagnosis among individuals summarizes the findings of the various methods and offers
between 15 and 39 years of age [2]. The efficient recommendations for future improvements in this field.
management of cerebral neoplasms relies on a meticulous
diagnosis, as distinct neoplasm types call for varying II. LITERATURE SURVEY
treatment tactics. The ramifications of a mistaken diagnosis
can be dire, emphasizing the utmost importance of precise
diagnostic approaches. 1.Advanced Brain Tumour Segmentation from MRI Images
written by Kavitha Angamuthu Rajasekaran and
Chellamuthu Chinna Gounder Et al.[1] The paper authored
focuses on the development and evaluation of advanced
methods for the automatic segmentation of brain tumors
from magnetic resonance imaging (MRI) images.The paper Thakur Et al.[6] This paper covers the various techniques that
likely starts with an introduction to the significance of brain are part of Image Processing and are used in finding brain
tumor segmentation in medical diagnosis and treatment tumors from MRI Images. A brief description of each
planning.The core of the paper describes the methodology technique is also provided like threshold OR fuzzy technique
or region based.Threshold Based Segmentation —Should be
proposed by the authors. This includes details about the
used when the contrast intensity is high Less susceptible to
image preprocessing steps, feature extraction techniques, noise than other images so can be used in conditions where
and the machine learning or deep learning architecture used no pre-filtering is done When quick analysis of the image is
for segmentation. Commonly used algorithms in this context to be done rather then exhaustive analysis Not able to extract
include Convolutional Neural Networks (CNNs) much information from the input MRI image and hence the
analysis may or not prove helpful Cannot be used for images
2. Identification of Brain Tumor using Image Processing with poor contrast or images with a lot of background and
Techniques written by Praveen Gamage Et al.[2] Abnormal foreground artifacts. Region Based - Watershed segmentation
growth of tissues in the brain which affect proper brain and region seed growing with well selected seed regions can
functions is considered as a brain tumor.MRI is mainly used be used to accurately extract object of interests like tumors
to get images of the human body and cancerous tissues form input MRI’s Noise may lead to undesired artifacts in
because of its high resolution and better quality images final result Susceptible to human error as seed region is
compared with other imaging technologies.Brain tumor manually selected
identifications through MRI images is a difficult task because
of the complexity of the brain.Two types of brain tumors have 7.A survey on brain tumor detection techniques for MR
been identified as Benign tumors and Malignant images written by Prabhjot Kaur Chahal,Shivani Goel Et
tumors.Benign tumors are less harmful than malignant al.[7] The article likely delves into the utilization of deep
tumors as malignant are fast developing and harmful while convolutional neural networks (CNNs) for precise brain
benign are slow growing and less harmful.The rest of the tumor detection. Deep CNNs are advanced machine learning
paper describes, background of identification of brain tumor models known for their exceptional image recognition
capabilities. In this context, the article explores how these
3.Sahaai, et al. (2022) [3] This study delves into the networks are employed to analyze medical images, such as
application of Convolutional Neural Networks (CNNs) in the brain MRIs or CT scans, to identify and precisely locate brain
classification of MRI and mammogram visuals into two tumors. The article may encompass various aspects,
categories: tumor and non-tumor. The investigation including data preprocessing, the design of the CNN
scrutinizes four distinct CNN models, namely DenseNet- architecture, training methodologies, and an assessment of
201, VGG-19, Xception, and Inception V3. The paper offers the CNN’s performance metrics such as sensitivity,
a comprehensive comparison of these CNN models, focusing specificity, and overall accuracy in detecting brain tumors.
on their accuracy and validation accuracy at the 10th, 20th, Furthermore, it is likely to discuss the potential implications
and 30th training epochs. DenseNet-201 emerges as the of such technology, particularly its role in early diagnosis and
frontrunner, surpassing the other models with an impressive optimizing treatment plans for individuals with brain tumors
91.94% accuracy after 30 epochs. The research centers on
evaluating the effectiveness of diverse CNN models for the 8 Ali pashaei, et al. (2022) [8] This study tackles the
categorization of brain tumors using MRI images. DenseNet- significant challenge of brain tumor classification through
201 establishes itself as the most precise model, achieving an the application of two distinct methods. The initial method
impressive accuracy rate of 91.9% following 30 training involves the use of a convolutional neural network (CNN)
epochs. Additional metrics further confirm its proficiency in equipped with multiple layers for feature extraction, resulting
distinguishing between tumor and non-tumor images.. in an accuracy rate of 81.09%. In the second method, CNN
handles the feature extraction, and KELM (Kernel Extreme
4.Filatov, et al. (2022) [4] This research centers on the Learning Machine) is harnessed for classification. This
application of machine learning techniques for the approach achieves a remarkable accuracy of 93.68%,
identification of brain tumors. The approach involves dataset surpassing the performance of the first method and previous
expansion through image augmentation, feature extraction findings, showcasing its enhanced accuracy in the
through pre-trained convolutional layers, and classification classification of brain tumors.
using a custom-designed artificial neural network. Image
augmentation is employed to address the limitation of a small 9 Brain Tumor Detection Using Deep-Learning Framework
dataset by generating image variations through translations, written by J HARSHINI and K GAYATRI. Et al.[9] This
rotations, scaling, shearing, and flipping. Two pre-trained article explores the application of a deep-learning framework
models, namely ResNet50 and EfficientNet, are utilized for
for the detection of brain tumors. Deep learning, a subset of
the classification task. EfficientNet models outshine
ResNet50, with EfficientNetB1 achieving the highest machine learning, involves training artificial neural networks
accuracy. The dataset consists of 7022 MRI visuals, which to perform specific tasks. In this context, the focus is on the
are divided into training and testing sets. identification of brain tumors in medical images, such as
MRIs or CT scans. Methodology: It may describe the deep-
5.Almadhoun, et al. (2022) [5] conclude a deep, easy, fast, learning framework used, which could include convolutional
and effective learning method to discover whether MRI neural networks (CNNs) or other architectures tailored for
images have a brain tumor, as it is a problem doctors face and image analysis. The choice of architecture often depends on
will help them make decisions easily, accurately and quickly. the complexity of the task and the available data

6 A Survey on Brain Tumor Detection Using Image 10 .MRI-based brain tumor detection using convolutional
Processing Techniques written by Luxit Kapoor Sanjeev deep learning methods and chosen machine learning
techniques written by Soheila Saeedi, Sorayya 14.Exposure to lead increases the risk of meningioma and
Rezayi,Sharareh R. Niakan Kalhori Et al.[10] .The article brain cancer written by Caoli Tang b, Jun Yu b, Shiyao Meng
may begin by highlighting the significance of early and c, Wanying Zhang a. Et al.[14] This association highlights
accurate detection of brain tumors, which can be crucial for the substantial health concerns associated with lead
timely treatment and improved patient outcomes. It likely exposure. Whether through contaminated water sources,
discusses the process of acquiring and preparing the MRI certain occupations, or environmental factors, exposure to
data for analysis. This could involve data augmentation, lead has been correlated with an increased likelihood of
normalization, and other preprocessing steps to ensure the developing meningioma, a type of brain tumor originating in
data is suitable for machine learning. The article explains the brain’s protective membranes, and various forms of brain
how the deep learning and machine learning models were cancer. The underlying mechanisms involve lead’s toxic
trained on the prepared data. It discusses the evaluation effects on the central nervous system, including the potential
metrics used to assess the models’ performance, which may for DNA damage and genetic mutations that contribute to
include accuracy, sensitivity, specificity, and the area under tumor development. Due to the cumulative and long-lasting
the receiver operating characteristic curve (AUC-ROC) impact of lead exposure, it is crucial to implement effective
public health measures and policies to reduce exposure and
11.Masood, et al. (2021) [11]This paper delves into the mitigate these serious health risks associated with brain
crucial aspect of brain tumor segmentation, underscoring its conditions
importance in enhancing patient survival rates and
simplifying surgical procedures. Given the drawbacks of 15.Ayesha Samreen, et al. (2020) [15] This research
manual segmentation, such as errors and time consumption, concentrates on the domain of deep learning, utilizing
there is a growing demand for automated methods. The paper Artificial Neural Networks (ANN) for comprehensive
presents an innovative strategy that employs the Mask RCNN analysis. Among ANN techniques, Convolutional Neural
model to achieve precise localization and segmentation of Networks (CNN) are widely acclaimed, with Python libraries
brain tumors. Noteworthy contributions include the model's such as TensorFlow and Keras proving instrumental in their
proficiency in addressing intricate structural variations, a application in the medical field. The paper unveils a
thorough quantitative assessment conducted on two distinct prototype algorithm intended for the detection of brain tumor
datasets, and the pioneering application of Mask RCNN in likelihood based on MRI images, using datasets from BRATS
the domain of brain tumor segmentation.Hari Mohan Rai, et and Kaggle. The algorithm incorporates image processing
al. (2021) techniques and CNN, and it outlines the experimental
methodology. The study introduces automatic algorithms for
[12] Computer-aided diagnosis of human brain tumor brain tumor detection via ANN, involving MRI image
through MRI: written by El-Dahshan ESA, Mohsen HM, preprocessing and post-processing, statistical feature
Revett K, Salem ABM Et al.[12] This article offers a analysis, and the use of artificial neural networks. These
comprehensive survey of computer-aided diagnosis (CAD) methods have exhibited noteworthy performance in brain
methods for detecting human brain tumors using MRI scans. tumor detection, surpassing existing state-of-the-art
Additionally, it introduces a new algorithm aimed at approaches. Future research should aim to enhance the
improving the accuracy and efficiency of brain tumor accuracy and efficiency of these algorithms.D C Febrianto,
diagnosis.Through an extensive literature review, the article et al. (2020) [16] uses Fuzzy C-Means and SVM methods to
examines various CAD techniques previously employed for segment brain parts with tumors, with
brain tumor diagnosis using MRI data. It explores the diverse 83.33% accuracy. The FLAIR method classifies MRI images
range of methods, including machine learning, deep learning, into normal and abnormal brains. CNN is proposed to solve
and image analysis, while shedding light on their respective data complexity by extracting features without removing
strengths and limitations.Introducing an innovative spatial information. However, it requires more data and time
approach, the article presents a detailed description of a novel compared to unsupervised learning approaches.
algorithm specifically developed to address the identified
challenges. This algorithm’s architecture, designed with 16.Review of MRI-based brain tumor image segmentation
unique features and enhancements, is outlined in depth.the using deep learning methods. written by Caoli Tang b, Jun Yu
authors summarize the critical findings from the survey and b, Shiyao Meng c, Wanying Zhang a. Et al.[16] This review
emphasize the importance of the new algorithm’s examines the use of deep learning techniques in segmenting
contribution to advancing the field of computer-aided brain brain tumor images from MRI scans. It covers deep learning
tumor diagnosis through MRI scans. methods, datasets, and evaluation metrics, addressing
challenges and clinical implications. This comprehensive
13.Mohammad Shahjahan,Detection of brain tumor et al. overview highlights the current state and future prospects of
(2021) [13] covers both conventional and deep learning MRI-based brain tumor image segmentation.
approaches for identifying brain tumors from MRI images.
The objectives of this research are to explore how various 17.Two-phase multi-model automatic brain tumour diagnosis
image processing techniques are applied on MRI images for system from magnetic resonance images using convolutional
the detection of brain tumors, compare the performance of neural networks. written by Abd-Ellah MK, Awad AI, Khalaf
existing image processing techniques, and propose an AAM, Hamed HFA. Et al.[17] This model uses
efficient technique for detecting brain tumors using MRI Convolutional Neural Networks” likely presents a novel
images through machine learning approaches. This research approach to automating brain tumor diagnosis from MRI
will help pathology experts provide proper treatment and scans through a two-phase process. In the first phase, it likely
improve the accuracy of machine learning algorithms in employs convolutional neural networks (CNNs) to perform
detecting brain tumors. initial image analysis and segmentation, effectively isolating
potential tumor regions from the MRI data. The second
phase likely involves the use of multiple models or Mohammad The CNN
modalities to further refine the diagnosis, possibly Omid Performance
Khairandish of
integrating additional clinical and radiological information..
(2020) Brain Tumor
Diagnosis
18.Amjad Rehman Khan, et al. (2020) [18] This research Based on
presents a comprehensive system designed for both the Machine
segmentation and classification of brain tumors using MRI learning
data, utilizing the VGG19 CNN model. To improve the Ayesha Human Brain Via Ann involving MRI technology
accuracy of classification, synthetic data augmentation Samreen Tumor
techniques are applied. The methodology encompasses the Detection
preprocessing of MRI data, the segmentation of brain tumors And
prevention
through Kmeans clustering, and their subsequent
classification into benign or malignant categories, employing
a fine-tuned VGG-19 CNN model. The system's performance Abhilash Brain Tumor SVM,Random Forest,K mean
Arjun Das Analysis segmentation
is evaluated using the BraTS 2015 dataset, demonstrating its
potential to assist radiologists and medical professionals in
the detection and categorization of brain tumors.
Comparative experiments reveal that this proposed approach
surpasses existing methods in terms of accuracy, particularly
in the realm of computer-aided diagnosis (CAD) systems.
III. PROPOSED WORK
19. Sabaa Ahmed Yahya and his team in 2020 [19], they
systematically explore deep learning (DL) techniques for the In this study, we suggest approach to enhance the
analysis of MRI brain medical images. The research identification and categorization of brain cancers in MRI
underscores the value of convolutional neural networks pictures. Our methodology is on the use of sophisticated
(CNNs) in a range of medical image processing tasks,
deep learning algorithms, primarily focusing on
including classification, identification, and segmentation.
The paper acknowledges the challenges associated with DL, Convolutional Neural Networks (CNNs). We will start by
particularly the constraints posed by limited data and labels, carefully choosing a good CNN architecture, keeping in
and discusses strategies such as data augmentation and mind the results and performance results from different
transfer learning to tackle these challenges. The study CNN models as mentioned in the body of existing
concludes by emphasizing the continued significance of DL research. The selected CNN architecture will be adjusted
in the analysis of MRI brain tumor medical images, and modified to meet the unique requirements of brain
emphasizing the need for ongoing efforts to address tumor identification.
associated issues in this field.
Preparing the data will be a crucial part of our procedure
20.Abhilash Arjun Das [20] discuss various research on brain in order to maximize the performance of the model. To
tumor classification using convolutional neuralnetworks. maintain uniformity, this will require resizing and
They use K-means segmentation, support vector machine standardizing MRI pictures. In addition, we will
(SVM), Random Forest investigate different methods of data augmentation,
including translation, rotation, scaling, and flipping, in
Table 1: Literature Overview order to lessen the impact of the small dataset and
strengthen the resilience of the model. An other important
Author Tittle Technology component of our suggested methodology is feature
Mahmoud Prediction
extraction. In order to improve the model's capacity to
cnn
Khaled Abd- Modelling of extract pertinent features from MRI pictures, we will look
Ellah, Ali Brain Tumor into the use of pre-trained convolutional layers and
Ismail Awad Detection investigate methods.In addition to increasing
Praveen Brain Tumor MRI,vgn19,cnn
classification accuracy, this will make it possible to
Gamage Classification
via comprehend the features that influence the model's
convulational decision-making process on a deeper level.
Neural
Network
Sahaai, Brain Tumor scrutinizes four distinct CNN
Classification models, namely DenseNet-201,
Using VGG-19, CONCLUSION
Machine To sum up, our study tackles the important problem of
Learning
Algorithm
enhancing the identification and categorization of brain
Amjad rehman MRI Brain K mean clustering,Cnn
tumors in magnetic resonance imaging. The timely and
khan (2021) tumor medical CAD precise identification of brain tumors plays a crucial role in
image analysis both patient outcomes and healthcare . This paper highlights
using deep the potential for deep learning—more specifically,
learning Convolutional Neural Networks—to revolutionize medical
techniques image processing and emphasizes the importance that these
networks play in this field.
The approach will not only help in developing system to
[15] Palvadi, Srinivas Kumar, and K. Suresh Joseph. "DETECTION AND
brain tumor detection but along with our approach we will CLASSIFICATION OF BRAIN TUMORS WITH ALZHEIMER
also be able to find type of tumor we have along with USING DEEP LEARNING AND CONVOLUTIONAL NEURAL
various sample of tumor dataset with us for training and NETWORKS.".
testing while developing this model of approach

REFERENCES [16] Işın, Ali, Cem Direkoğlu, and Melike Şah. "Review of MRI-
based brain tumor image segmentation using deep learning
[1] Rajasekaran, Kavitha Angamuthu, and Chellamuthu Chinna Gounder. methods." Procedia Computer Science 102 (2016): 317-324.
"Advanced brain tumour segmentation from mri images." Basic
physical principles and clinical applications, high-resolution
neuroimaging (2018) [17] Abd-Ellah, Mahmoud Khaled, et al. "Two-phase multi-model
automatic brain tumour diagnosis system from magnetic resonance
images using convolutional neural networks." EURASIP Journal on
[2] Akshaya, T. A. M., et al. "Identification of Brain Tumor on Mri images Image and Video Processing 2018.1 (2018): 1-10.
with and without Segmentation using DL Techniques." E3S Web of
Conferences. Vol. 399. EDP Sciences, 2023.
[18] Rehman A, Khan MA, Saba T, Mehmood Z, Tariq U, Ayesha N.
Microscopic brain tumor detection and classification using 3D CNN
[3] Zhu, Ziquan, Shui-Hua Wang, and Yu-Dong Zhang. "A survey of
convolutional neural network in breast cancer." Computer modeling in and feature selection architecture.
engineering & sciences: CMES 136.3 (2023): 2127.

[4] Arabahmadi, Mahsa, Reza Farahbakhsh, and Javad Rezazadeh. "Deep

learning for smart Healthcare—A survey on brain tumor detection from
medical imaging." Sensors 22.5 (2022): 1960.

[5] Almadhoun, Hamza Rafiq, and Samy S. Abu-Naser. "Detection of brain

tumor using deep learning." (2022).

[6] Kapoor, Luxit, and Sanjeev Thakur. "A survey on brain tumor detection
using image processing techniques." 2017 7th international conference
on cloud computing, data science & engineering-confluence. IEEE,
2017.

[7] Chahal, Prabhjot Kaur, Shreelekha Pandey, and Shivani Goel. "A
survey on brain tumor detection techniques for MR
images." Multimedia Tools and Applications 79 (2020): 21771-21814.

[8] Dehkordi, Amin Abdollahi, et al. "Brain Tumor Detection and

Classification Using a New Evolutionary Convolutional Neural
Network." arXiv preprint arXiv:2204.12297 (2022). Ankit Ghosh, &
Alok Kole. (2021).

[9] Sethi, Rachna, et al. "2nd International Conference on Inventive

Research in Computing Applications (ICIRCA 2020)." Evaluation 1

[10] Saeedi, Soheila, et al. "MRI-based brain tumor detection using

convolutional deep learning methods and chosen machine learning
techniques." BMC Medical Informatics and Decision Making 23.1
(2023)

[11] Das, Surajit, and Rajat Subhra Goswami. "Review, Limitations, and
future prospects of neural network approaches for brain tumor
classification." Multimedia Tools and Applications (2023): 1-43.

[12] El-Dahshan, El-Sayed A., et al. "Computer-aided diagnosis of human

brain tumor through MRI: A survey and a new algorithm." Expert
systems with Applications 41.11 (2014): 5526-5545.

[13] MAJIB, MOHAMMAD SHAHJAHAN. DETECTION OF BRAIN

TUMORS FROM MRI IMAGES USING MACHINE LEARNING
APPROACHES. Diss. DEPARTMENT OF COMPUTER SCIENCE
AND ENGINEERING, 2021.

[14] Meng, Yu, et al. "Exposure to lead increases the risk of meningioma and
brain cancer: A meta-analysis." Journal of Trace Elements in Medicine
and Biology 60 (2020): 126474.