NEUROSENTINEL PRODIGY (Brain Tumor) using CNN

VIVA-Tech International Journal for Research and Innovation
ISSN(Online): 2581-7280
Volume 1, Issue 6 (2023)
Article No. 11
PP 01-13
1
www.viva-technology.org/New/IJRI
10
NEUROSENTINEL PRODIGY
Shaad Shaikh1
, Vansh Shah2
, Deven Randive3
, Prof. Saniket
Kudoo4
1
(Department Computer Engineering, Mumbai University, MUMBAI)
2
3
4
(Professor, Department Computer Engineering, Mumbai University, MUMBAI)
Abstract: Brain tumor detection is a crucial task in the realm of medical diagnostics, bearing significant
implications for patient care and outcomes. This research paper embarks on a comprehensive exploration of the
development and deployment of an advanced brain tumor detection system. The methodological framework is
multifaceted, commencing with the assembly of a diverse and extensive dataset of brain imaging scans. Subsequently,
the data undergoes rigorous preprocessing, including noise reduction and image enhancement, to optimize the
quality and fidelity of the scans. The heart of the system lies in the utilization of deep learning, particularly a
convolutional neural network (CNN), which leverages the robust features extracted from the preprocessed data to
distinguish between brain scans indicative of tumors and those that are not. Model training is augmented by the
introduction of a validation set, allowing for finetuning to achieve optimal performance. Testing the trained model on
an entirely separate and previously unseen dataset substantiates its real-world utility, providing critical insights into
its robustness and accuracy. The practical implementation of the system involves seamless integration into a real-
time processing platform, enabling rapid analysis of incoming brain imaging data. This operational phase includes
the establishment of predefined thresholds, effectively reducing false alarms and ensuring that only the most
probable cases are flagged for review by medical professionals.
Keywords – Brain Tumor, CNN, Medical.
I. INTRODUCTION
The early and accurate detection of brain tumors remains a critical challenge in the field of medical diagnostics
and healthcare. Brain tumors are abnormal growths of cells within the brain or the surrounding structures, and
they can have severe and life-threatening consequences if not detected and treated promptly. The timely
diagnosis of brain tumors is essential to initiate appropriate medical interventions and enhance the chances of a
successful outcome for patients. Brain tumor detection has witnessed remarkable advancements in recent years.
The integration of medical imaging techniques, computational algorithms, and machine learning methods has
significantly improved the precision, efficiency, and speed of brain tumor diagnosis. Various methodologies
and approaches employed in the domain of brain tumor detection, with a focus on the innovative techniques
that have emerged as powerful tools in this field. The significance of early brain tumor detection cannot be
overstated. Brain tumors can manifest with a wide range of symptoms, some of which are subtle and easily
mistaken for other medical conditions. As a result, patients may go undiagnosed for extended periods, leading
to delayed treatment and compromised outcomes. Furthermore, the complexity of the human brain and the
intricate nature of brain tumors pose challenges for accurate detection and classification. These challenges
necessitate the development of advanced diagnostic systems that can assist healthcare professionals in making
informed decisions and offering personalized treatment strategies.

Article No. 11
PP 01-13
2
10
II. REVIEW OF LITERATURE SURVEY
The following chapter is a literature survey of the previous research papers and research which gives
the detailed information about the previous system along with its advantages and disadvantages.
Tariq Sadad, Amjad Rehman, Asim Muni,r Tanzila Saba, Usman Tariq, Noor Ayesha, Rashid Abbasi
et.al [1] In this research paper, the authors address the critical issue of brain tumor detection and classification
using advanced deep learning techniques applied to MRI slices. The study leverages transfer learning through
freeze and fine-tune processes to extract meaningful features from the MRI data. For tumor detection, they
employ the Unet architecture with ResNet50 as a backbone, achieving an impressive Intersection over Union
(IoU) score of 0.9504, showcasing the model's accuracy in delineating tumor regions. Furthermore, the paper
extends its focus to multi-classification, distinguishing between different types of brain tumors, using the
powerful NASNet architecture. The NASNet model outperforms other deep learning architectures with a
remarkable classification accuracy of 99.6%. The study acknowledges the need for future research to explore
architectures with reduced computational complexity, promising to further enhance the field of automated brain
tumor detection and classification. Overall, this paper demonstrates the potential of deep learning and transfer
learning in improving the accuracy and efficiency of brain tumor diagnosis, offering valuable insights for
medical professionals and researchers in the field.
Pallavi Tiwari, Bhaskar Pant, Mahmoud M. Elarabawy, Mohammed Abd-Elnaby, Noor Mohd, Gaurav
Dhiman and Subhash Sharma et.al [2] This research paper introduces a groundbreaking method for the
automated multiclass classification of brain tumors in MRI images, leveraging a sophisticated deep
Convolutional Neural Network (CNN) architecture. The dataset under scrutiny is comprehensive, containing a
total of 3,264 MRI images, thoughtfully divided into four distinct classes: glioma, meningioma, no tumor, and
pituitary tumors. The heart of the study lies in the proposed CNN model, which encompasses a total of six
layers, intricately combining convolutional layers for feature extraction, batch normalization to enhance
autonomous learning, activation functions for non-linearity, pooling layers for dimensionality reduction,
dropout layers for regularization, and a fully connected layer for classification. The crowning achievement of
this model is its remarkable accuracy, clocking in at an impressive 99%, while concurrently maintaining a low
loss of 0.0504 across 30 epochs of training. An indepth comparative analysis against existing models solidifies
the superiority of the proposed approach. Additionally, the study visually articulates model performance
through a comprehensive confusion matrix and a detailed classification report. This research not only
underscores the potency of deep CNNs in the realm of automated medical image classification but also
underscores their efficacy in situations characterized by limited training data and minimal preprocessing
requirements. As a harbinger of the future, the study aims to extend its classification capabilities and boost
accuracy, setting the stage for further advancements in AIdriven diagnostic tools. In sum, this research is a
significant stride towards revolutionizing the diagnosis and treatment planning of brain tumors, offering
promising avenues for enhancing patient care and outcomes in clinical practice.
Tanzila Saba, Ahmed Sameh Mohamed, Mohammad El-Affendi, Javeria Amin, Muhammad Sharif
et.al [3] The research paper presents a comprehensive methodology for the accurate diagnosis and classification
of brain tumors, with a primary focus on gliomas, utilizing medical imaging data. The authors emphasize the
critical need for early detection of brain tumors, given the potential life-threatening consequences of rapid
tumor growth and pressure on surrounding healthy brain tissue. The proposed approach combines advanced
image processing techniques, deep learning, and traditional handcrafted features to achieve precise tumor
segmentation and subsequent classification. The segmentation process employs the GrabCut method, which
initially converts RGB images into single-channel representations and iteratively refines tumor boundaries
based on seed points and pixel similarity thresholds. Deep learning features are extracted using a fine-tuned
VGG-19 model, while handcrafted features, such as Local Binary Pattern (LBP) and Histogram of Oriented
Gradients (HOG), are also employed. These features are optimized through entropy-based techniques and fused
into a single feature vector for classification. The research evaluates the methodology on three benchmark
datasets: BRATS 2015, 2016, and 2017, which include varying numbers of high-grade and low-grade glioma
cases. Results demonstrate high accuracy in tumor classification across multiple classifiers, including Support
Vector Machines (SVM), Logistic Regression (LGR), and an ensemble classifier. Additionally, the method is
successful in distinguishing between high and low-grade tumors. Comparative analysis reveals that the
proposed approach surpasses existing state-ofthe-art methods in brain tumor classification. The research
concludes by summarizing its strengths, highlighting the achievements of the proposed approach in terms of
accuracy and Dice Similarity Coefficient (DSC) on the tested datasets. Overall, this paper contributes to the
field of medical image analysis by offering a robust and effective solution for the early diagnosis and

Article No. 11
PP 01-13
3
10
classification of brain tumors, potentially leading to improved patient outcomes in the context of this critical
health concern.
Md Khairul Islam, Md Shahin Ali, Md Sipon Miah, Md Mahbubur Rahman, Md Shahariar Alam,
Mohammad Amzad Hossain et.al [4] The research paper presents superpixels, PCA & K-means scheme for
brain tumor detection in magnetic resonance (MR) images. The study begins by emphasizing the importance of
accurate tumor detection in medical imaging, particularly in MR images, which are often complex and require
preprocessing for noise reduction. The proposed scheme comprises several key components, including image
preprocessing, feature extraction using superpixels and principal component analysis (PCA), and tumor
segmentation using template-based K-means clustering. The authors conducted experiments using a dataset of
40 MR images, showcasing the effectiveness of their scheme. They achieved an impressive accuracy of 95.0%,
sensitivity of 97.36%, and specificity of 100%, surpassing existing detection methods. Superpixels and PCA
were found to be instrumental in dimensionality reduction and simplification of MR images, facilitating
accurate tumor detection. The proposed scheme also demonstrated fast execution times, making it practical for
clinical applications. However, the study acknowledges limitations such as the use of a small dataset. Future
research will address these limitations, aiming to enhance the accuracy of detection and classification, stage
identification of tumors, and compatibility with deep learning systems for broader applicability in various
radiological imaging techniques. Overall, the paper presents a promising approach to improve brain tumor
detection in MR images and highlights avenues for future research and development.
Chirodip Lodh Choudhury, Chandrakanta Mahanty, Raghvendra Kumar et.al [5] The research paper
presents a novel approach to brain tumor detection using a Convolutional Neural Network (CNN) applied to
MRI images. The study addresses the critical need for early and accurate detection of brain tumors, which can
significantly impact patient outcomes and treatment options. The proposed CNN architecture consists of three
layers, and the model achieved an impressive accuracy rate of 96.08% with an F-score of 97.3. The research
highlights the power of machine learning in medical diagnostics, particularly in the field of neuro-oncology, by
significantly outperforming traditional manual diagnosis methods. The CNN model leverages its ability to learn
hierarchical features from medical images, from basic attributes like edges to more complex features. Various
activation functions, including Rectified Linear Unit (ReLU), Hyperbolic Tangent (Tanh), and Sigmoid, are
utilized for efficient learning, and the training process is optimized using the ADAM optimizer. The results, as
depicted in the confusion matrix, illustrate a minimal error rate of 2.98%. In conclusion, this research
underscores the potential of diagnostic machine learning applications and predictive treatment in healthcare and
points to future research avenues, particularly the application of "neutrosophical principles" for brain tumor
detection using CNNs, promising advancements in the field of medical image analysis and diagnosis.
Sarmad Maqsood, Robertas Damaševi cius and Rytis Maskeliunas et.al [6] The research paper presents
a comprehensive approach to the detection and classification of brain tumors using medical imaging, addressing
critical challenges in the field. Brain cancer, a leading global cause of mortality, necessitates early and precise
diagnosis, which can be facilitated through magnetic resonance imaging (MRI). Manual detection is time-
consuming and error-prone, motivating the development of an automated computer-aided diagnosis (CAD)
method. The proposed framework comprises contrast enhancement, image segmentation, feature extraction,
feature selection, and classification. Linear contrast stretching enhances image contrast, followed by a custom
17-layered CNN for tumor segmentation. Feature extraction utilizes a modified MobileNetV2 architecture with
transfer learning. Entropy-based feature selection refines feature sets, and multiclass support vector machines
(M-SVM) perform tumor classification, distinguishing meningioma, glioma, and pituitary images.
Experimental evaluations conducted on BraTS 2018 and figshare datasets reveal exceptional performance, with
classification accuracy rates reaching 97.47% and 98.92%, respectively, outperforming existing methods.
Notably, the use of gradient-weighted Class Activation Mapping (Grad-CAM) offers visual insights into
regions influencing tumor classifications. Limitations include a focus on 2-D MRI images and a slightly time-
consuming feature selection process. Future work aims to extend the methodology to 3-D imaging and address
time efficiency concerns. In conclusion, the proposed CAD system demonstrates substantial promise in
improving brain tumor diagnosis and classification, providing enhanced accuracy, automatic feature extraction,
reduced computational time, and effective feature selection, representing a significant advancement in medical
image analysis.
Mesut TOĞAÇAR, Burhan ERGEN, Zafer CÖMERT et.al [7] The paper presents BrainMRNet, a
novel convolutional neural network model designed for the detection of brain tumors in magnetic resonance
images. Brain tumors can have life-threatening consequences, and their early diagnosis is of paramount
importance. The proposed model integrates attention modules, a hypercolumn technique, and residual blocks to
enhance the detection accuracy. In a two-step experimental study, the BrainMRNet model was compared with
pre-trained CNN models, including AlexNet, GoogleNet, and VGG-16. In the first step, the pre-trained CNNs
achieved classification accuracies ranging from 84.48% to 89.66%. However, in the second step, BrainMRNet
outperformed them all with a classification accuracy of 96.05%, along with high sensitivity and specificity. The

Article No. 11
PP 01-13
4
10
attention modules allowed the model to focus on relevant areas of MR images, the hypercolumn technique
retained features from different layers, and the residual blocks minimized issues related to network depth. These
components contributed to the improved performance of BrainMRNet. The study compared its results with
previous works, demonstrating its superiority. Despite the dataset's low resolution, the model's accuracy
exceeded 96%, showing promise for early diagnosis of brain tumors. The open-source code for BrainMRNet is
available for further development and applications in medical image analysis. The study concludes by
suggesting the potential use of the proposed model in various medical image analysis applications and fields in
future research.
Shtwai Alsubai, Habib Ullah Khan, Abdullah Alqahtani, Mohemmed Sha, Sidra Abbas and Uzma
Ghulam Mohammad et.al [8] The proposed methodology for brain tumor detection in MRI images integrates
advanced deep learning techniques, such as Convolutional Neural Networks (CNN) and a hybrid CNN-Long
Short-Term Memory (LSTM) model. This methodology begins with the preprocessing of MRI images,
including resizing, cropping using extreme point calculation, and bicubic interpolation. The dataset is divided
into training and validation sets. The CNN is employed for feature extraction, and the CNN-LSTM hybrid
model acts as the classifier. Both models achieve high accuracy, with CNN-LSTM outperforming CNN.
Training accuracy for CNN is 99.4%, and validation accuracy is 98.3%, while training loss is 0.007, and
validation loss is 0.113. In contrast, CNN-LSTM exhibits superior performance, with training and validation
accuracies of 99.8% and 98.5%, respectively, and training and validation losses of 0.010 and 0.103. The results
of the proposed technique are impressive, with CNN achieving 98.6% accuracy, 98.5% precision, 98.6% recall,
and an F1- measure of 98.4%. The hybrid CNN-LSTM model surpasses these metrics with an accuracy of
99.1%, precision of 98.8%, recall of 98.9%, and an F1-measure of 99.0%. Graphical representations of the
performance illustrate CNN-LSTM's superiority. Comparative analysis reveals that the proposed model
outperforms existing techniques, further confirming its effectiveness. This methodology demonstrates the
potential of deep learning and the CNNLSTM hybrid model for accurate and efficient brain tumor detection in
MRI images, offering promising prospects for improved healthcare diagnosis and treatment in the future.
Wadhah Ayadi, Wajdi Elhamzi, Imen Charf, Mohamed Atri et.al [9] In this research paper, a novel
and robust approach for the classification of brain tumors in MRI images is presented. The methodology
employs deep Convolutional Neural Networks (CNNs) for automated diagnosis, offering a solution that
minimizes preprocessing requirements. The study encompasses the evaluation of the proposed model using
three distinct datasets, showcasing its adaptability and effectiveness. Extensive data augmentation techniques,
including rotation, flipping, Gaussian blur, and sharpening, were applied to enhance the model's performance.
The results are evaluated using various metrics, such as accuracy, sensitivity, specificity, precision, and F1-
score, providing a more comprehensive assessment than accuracy alone. In the experiments, the proposed CNN
model consistently outperformed previous works across the different datasets, demonstrating its potential for
real-world medical applications. Future directions for this research include exploring ensemble methods,
improving model interpretability, investigating transfer learning, optimizing for efficiency, expanding datasets,
addressing deployment challenges, and collaborating with medical experts to further refine and adapt the model
for clinical use. This research offers a promising avenue for enhancing the accuracy and efficiency of brain
tumor diagnosis through medical image analysis.
Neelum Noreen, Sellappan Palaniappan, Abdul Qayyum, Iftikhar Ahmad, Muhammad Imran, and
Muhammad Shoaib et.al [10] The research paper presents a comprehensive study on the application of deep
learning models for the early detection and classification of brain tumors using magnetic resonance imaging
(MRI) data. Two distinct scenarios are explored in this study, utilizing pre-trained deep learning models,
namely DenseNet201 and Inception-v3, to extract features from MRI images for brain tumor classification. The
dataset used comprises 3,064 T1-weighted contrast MR images of three different types of brain tumors:
meningioma, glioma, and pituitary tumors. The study introduces a feature concatenation approach, combining
features from different layers or blocks of the pre-trained models, followed by softmax classification. Results
indicate remarkable performance, with an accuracy of 99.51% achieved using the DenseNet201-based
ensemble method, outperforming existing approaches. The paper discusses the challenges of traditional manual
feature extraction methods and highlights the capability of deep learning models, particularly convolutional
neural networks (CNNs), to automatically extract relevant features from medical images. It underscores the
need for efficient and reproducible computer-aided diagnosis tools to process large-scale medical datasets,
emphasizing the complexity of brain tumor classification due to variations in tumor location, shape, size, and
intensity. The research concludes that deep learning models provide a powerful and promising approach for
brain tumor identification, with future directions including fine-tuning, data augmentation, scratch-based
models, and ensemble methods to further enhance classification accuracy. This study contributes significantly
to the field of medical image analysis, offering an automated and accurate solution for early brain tumor
detection and classification through the integration of advanced deep learning techniques.
Joshi Manisha, Umadevi, Akshitha Raj B N et.al [11] this research paper presents a approach to the

Article No. 11
PP 01-13
5
10
multi-classification of brain tumors using convolutional neural networks (CNNs). Brain tumor diagnosis
traditionally relies on invasive, time-consuming, and error-prone histopathological analysis of biopsy
specimens. The paper addresses this challenge by introducing three distinct CNN models, each tailored to a
specific classification task. The first model, Classification-1, achieves an impressive 99.33% accuracy in
detecting brain tumors. The second model, Classification-2, classifies brain tumors into five types (glioma,
meningioma, pituitary, normal brain, and metastatic) with an accuracy of 92.66%. The third model,
Classification-3, grades glioma tumors into three categories (Grade II, Grade III, and Grade IV) with an
accuracy of 98.14%. Notably, the majority of hyper-parameters for these models are automatically tuned using
a grid search optimization algorithm. The paper compares the performance of these CNN models with popular
pre-trained networks, consistently demonstrating their superiority. Additionally, the study highlights the
importance of architectural engineering in deep learning, moving away from traditional feature engineering.
This research stands as a pioneering contribution, being the first to address multi-classification of brain tumor
MRI images using CNNs with extensive hyper-parameter optimization. In summary, the proposed CNN models
exhibit exceptional accuracy in brain tumor detection, classification, and grading, showcasing their potential to
aid early diagnosis and support medical professionals, with automated hyper-parameter tuning further
enhancing their robustness and effectiveness.
Francisco Javier Díaz-Pernas, Mario Martínez-Zarzuela, Míriam Antón-Rodríguez and David
González-Ortega et.al [12] This research paper introduces a novel and fully automated approach for the
segmentation and classification of brain tumors in MRI images. The proposed method utilizes a Multiscale
Convolutional Neural Network (CNN) architecture, inspired by the human visual system, to process MRI slices
at three different spatial scales. Unlike previous methods, this approach does not require preprocessing to
remove skull or vertebral column parts from the input images. The CNN processes each pixel within a sliding
window, classifying it into one of four categories: healthy region, meningioma tumor, glioma tumor, or
pituitary tumor. The network consists of three pathways, each with varying kernel sizes to extract features at
different scales, and a fully connected stage for classification. Data augmentation using elastic transformation is
employed to prevent overfitting. The method is evaluated on a dataset of 3064 MRI slices from 233 patients,
covering three tumor types. Performance is compared to classical machine learning and deep learning methods,
and the proposed approach achieves a remarkable tumor classification accuracy of 0.973, outperforming all
other methods. Segmentation results are also impressive, with an average Dice index of 0.828, Sensitivity of
0.940, and pttas of 0.967. Even though some false positives occur, particularly in meningioma cases, the
method demonstrates robustness and competitiveness in brain tumor analysis. In conclusion, this fully
automatic CNN-based approach presents a powerful tool for brain tumor segmentation and classification,
offering great potential for assisting medical professionals in diagnosing brain tumors and showing promise for
application in other domains like satellite image analysis in future research endeavors.
Nyoman Abiwinanda, Muhammad Hanif, S. Tafwida Hesaputra, Astri Handayani, and Tati Rajab
Mengko et.al [13] In this research paper, a Convolutional Neural Network (CNN) was developed and evaluated
for the automated classification of common brain tumor types, specifically Glioma, Meningioma, and Pituitary
tumors, using T-1 weighted CE-MRI images. Notably, the study aimed to streamline the classification process
by eliminating the need for region-based pre-processing steps. The authors explored multiple CNN
architectures, ultimately identifying an optimal architecture, referred to as "architecture 2," which consisted of
two convolution layers with ReLU activation functions and max-pooling, followed by a hidden layer with 64
neurons. Architecture 2 exhibited a consistent decrease in validation loss during training, leading to the highest
validation accuracy of 84.19%, while the training accuracy reached 98.51%. Importantly, these results were
competitive with conventional algorithms that relied on region-based pre-processing, which achieved
accuracies ranging from 71.39% to 94.68%. The paper also discussed the potential for future improvements,
including the integration of color balancing techniques to enhance accuracy in textured MRI pixels. Overall, the
study highlighted the promise of CNNs as supportive tools for medical professionals in the classification of
brain tumors, offering the potential to simplify and expedite the diagnostic process while maintaining high
levels of accuracy and patient care.
Jayshree Ghorpade Aher, Abhishek Patil, Yudhishthir Deshpande, Eeshan Phatak et.al [14] has
proposed on A proposed framework for prediction of pulse, based on the effect of surya namaskar on different
prakruti at different prahars of the day. The study emphasizes the significance of Nadi-parikshan, a crucial
diagnostic technique in Ayurveda based on wrist pulse analysis. Introducing a novel framework, it aims to
predict how performing Surya Namaskar influences wrist pulse signals, considering different Prakruti types and
varied times of the day (Prahar). By integrating questionnaire data and leveraging machine learning algorithms,
this approach promises to offer tailored insights into holistic health, thereby contributing to a more personalized
understanding of one's well-being and daily life cycle.
Masoumeh Siar, Mohammad Teshnehlab et.al [15] This research paper presents a novel approach to
brain tumor detection using a combination of feature extraction algorithms and Convolutional Neural Networks

Article No. 11
PP 01-13
6
10
(CNNs). The study utilizes a dataset of MRI images from 153 patients, comprising both normal and brain tumor
cases, and introduces an innovative methodology for tumor detection. The CNNs are shown to be effective in
automating feature extraction from medical images, enhancing the accuracy of tumor identification. The paper
employs various classifiers, including the Softmax Fully Connected layer, Radial Basis Function (RBF)
classifier, and Decision Tree (DT) classifier, to evaluate network performance. Results reveal that the Softmax
classifier in the CNN achieved an accuracy of 98.67% in image categorization. To further enhance network
accuracy, a new method is proposed, combining feature extraction with the CNN. The combined approach
demonstrates substantial improvement, achieving an accuracy of 99.12% on test data. This heightened accuracy
is expected to have a significant clinical impact by aiding physicians in making more precise tumor diagnoses
and consequently improving patient care and treatment outcomes. In summary, this paper provides a promising
methodology for early and accurate brain tumor detection, leveraging the capabilities of CNNs and feature
extraction techniques, ultimately contributing to enhanced medical accuracy and patient care.

Article No. 11
PP 01-13
7
10
III. ANALYSIS
Analysis table summarizes the research papers on the Face Recognition. Below is a detailed description
of various algorithms used in research papers.
Table 1: Analysis Table
Title Summary Advantages TechStack
Brain tumor
detection and
multi-
classification
using advanced
deep learning
techniques. [1]
This article explores the
use of deep learning
techniques for brain
tumor detection and
classification from MRI
slices. It employs
transfer learning
processes, including
freeze and fine-tune, to
extract features. The
Unet architecture with
ResNet50 achieves high
accuracy (IoU of
0.9504) in tumor
detection.
The research achieves a
high accuracy rate of
99.6% in brain tumor
classification, which can
significantly enhance
medical diagnosis and
treatment planning.
The paper lacks an in-
depth discussion of
potential limitations,
data biases, and
computational resource
requirements,
potentially limiting its
real-world applicability
CNN Based
Multiclass Brain
Tumor Detection
Using Medical
Imaging [2]
This research paper
introduces a novel deep
Convolutional Neural
Network (CNN) model
for automated brain
tumor classification in
MRI images, achieving
an impressive 99%
accuracy. The dataset
comprises over 3,000
MRI images across four
classes: glioma,
meningioma, no tumor,
and pituitary tumors.
The proposed deep
Network (CNN) model
achieves an impressive
99% accuracy in
classifying brain tumors
from MRI images,
providing a highly
accurate and efficient
diagnostic tool for
medical professionals.
Despite its high
accuracy, the model's
success heavily relies on
the availability of a
substantial and well-
labeled dataset, which
may not always be
accessible in real-world
clinical settings.
Brain tumor
detection using
fusion of hand
crafted and deep
learning features
[3]
The Paper presents an
integrated approach for
the accurate diagnosis
of brain tumors,
focusing on gliomas,
using medical imaging.
This methodology
combines advanced
image segmentation
techniques, deep
learning, and traditional
handcrafted features to
achieve precise tumor
delineation and
classification.
The proposed
methodology offers
enhanced accuracy in
brain tumor diagnosis
through the fusion of
deep learning and
handcrafted features,
leading to improved
patient outcomes.
The methodology's
performance could be
sensitive to variations in
the quality of input
imaging data, which
may limit its robustness
in real-world clinical
scenarios with diverse
data sources.

Article No. 11
PP 01-13
8
10
Brain tumor
detection in MR
image using
superpixels,
principal
component
analysis and
template based K-
means clustering
algorithm [4]
This research paper
introduces a novel
scheme for brain tumor
detection in magnetic
resonance (MR) images,
utilizing superpixels,
PCA, and template-
based K-means
clustering. The
proposed scheme
demonstrates superior
accuracy, achieving
95.0% accuracy,
97.36% sensitivity, and
100% specificity
The proposed scheme
effectively reduces the
dimensionality and
complexity of MR
images through
superpixels and PCA,
enabling more efficient
feature extraction and
tumor detection.
The study employs a
relatively small dataset
and does not incorporate
the 2019 WHO
international
classification of
diseases, limiting its
real-world clinical
applicability.
Brain Tumor
Detection and
Classification
Using
Convolutional
Neural Network
and Deep Neural
Network [5]
This research paper
introduces a
Network (CNN) based
approach for brain
tumor detection using
MRI images, achieving
an impressive accuracy
of 96.08% and an F-
score of 97.3%. The
study emphasizes the
potential of machine
learning in enhancing
medical diagnostics
The CNN-based model's
ability to automatically
extract and learn
complex features from
MRI images improves
diagnostic accuracy and
reduces the subjectivity
associated with manual
interpretation,
enhancing the efficiency
of brain tumor detection.
The model's
performance may be
limited by the
availability and quality
of MRI data, and it may
require substantial
computational resources
for training and
inference.
Multi-Modal
Brain Tumor
Detection Using
Deep Neural
Network and
Multiclass SVM
[6]
The proposed method
combines contrast
enhancement, custom
CNN segmentation,
MobileNetV2 feature
extraction, and entropy-
based feature selection,
achieving outstanding
classification accuracy
rates of 97.47% and
98.92% on two datasets.
The proposed automated
brain tumor detection
system offers enhanced
accuracy, reducing
reliance on error-prone
manual diagnosis by
radiologists.
The method's current
limitation is its
applicability to 2-D
MRI images, potentially
limiting its effectiveness
in 3-D imaging
scenarios.
BrainMRNet:
Brain Tumor
Detection using
Magnetic
Resonance Images
with a Novel
Convolutional
Neural Network
Model [7]
This study introduces
BrainMRNet, a novel
convolutional neural
network, for the
detection and
classification of brain
tumors using MRI
images.
BrainMRNet's advanta
ge lies in its superior
performance, achieving
a 96.05% classification
success rate, surpassing
existing models.
Python 3.6, Python
libraries, Jupyter
Notebook, and the Keras
framework for neural
network development

Article No. 11
PP 01-13
9
10
Ensemble deep
learning for brain
tumor detection
[8]
This paper proposes a
hybrid deep learning
model (CNN-LSTM)
for brain tumor
classification using MRI
images, achieving high
accuracy and precision
(99.1% and 98.8%,
respectively). The study
addresses the challenges
of brain tumor detection
and emphasizes the role
of deep learning in early
diagnosis, contributing
to improved patient
outcomes.
The CNN-LSTM model
offers a remarkable
99.1% accuracy,
enabling early brain
tumor detection and
better patient outcomes.
Networks (CNN) and
Long Short-Term
Memory (LSTM) for
brain tumor detection
Deep CNN for
Brain Tumor
Classification [9]
Automated computer
assisted diagnosis
(CAD) systems are
needed due to the
complexity and volume
of data. The proposed
CNN-based model
demonstrates strong
performance in brain
tumor classification,
even with limited
training data, addressing
the limitations of
previous methods.
The advantage of using
Networks (CNNs) in
brain tumor
classification is their
ability to automatically
extract meaningful
features from MRI
images, reducing the
need for manual feature
engineering and
improving accuracy.
Includes deep
Networks (CNNs) for
image analysis and
classification,
complemented by data
augmentation
techniques, with a focus
on medical imaging
datasets.
A Deep Learning
Model Based on
Concatenation
Approach for the
Diagnosis of
Brain Tumor [10]
Use of pre-trained deep
learning models,
DenseNet201 and
Inception-v3, for the
tumors in MRI images,
achieving accuracy rates
of up to 99.51% through
feature concatenation
and softmax
classification.
Accuracy achieved
through deep learning
models, offering an
automated and efficient
solution for brain tumor
classification in MRI
images.
DensNet201 and
Inception-v3,
implemented in Keras
with the TensorFlow
backend for the
tumors using MRI
images.
Multi-
Classification of
Brain Tumor MRI
Images Using
Deep
Convolutional
Neural Network
with Fully
Optimized
Framework [11]
Paper presents three
dedicated convolutional
neural network (CNN)
models for brain tumor
classification, achieving
ac curacy level
(98.14%). Through
automatic hyper
parameter tuning via
grid search
optimization, these
models consistently
outperform pre-trained
networks, offering high
The paper's automatic
hyper-parameter tuning
enhances the CNN
models' performance,
ensuring robust and
accurate brain tumor
classification.
(CNNs) for image
classification, hyper-
parameter optimization
techniques like grid
search, and standard
performance evaluation
metrics

Article No. 11
PP 01-13
10
10
accuracy solutions for
brain tumor detection,
classification.
A Deep Learning
Approach for
Brain Tumor
Classification and
Segmentation
Using a
Multiscale
Convolutional
Neural Network
[12]
This research paper
presents a fully
automated method for
brain tumor
segmentation and
classification using a
Multiscale
Network (CNN)
architecture. The
proposed CNN achieves
a tumor classification
accuracy of 0.973 on a
dataset containing three
tumor types.
The proposed
Multiscale CNN offers
fully automated brain
tumor segmentation and
classification, reducing
the need for manual
intervention in medical
image analysis
PyTorch, image
processing libraries, and
possibly hardware
accelerators such as
Nvidia GPUs for model
training.
Classification of
Brain Tumor
Using
Convolutional
Neural Network
[13]
Utilizing a CNN for
image classification
with an impressive
accuracy of 98% on a
dataset comprising 330
brain images. And
introduces an Watershed
Algorithm for precise
segmentation,
demonstrates its
effectiveness in
distinguishing tumor
regions from normal
brain tissue
The integrated approach
of CNN based image
classification and the
Watershed Algorithm
for segmentation offers
high accuracy in brain
tumor detection and
localization.
Includes Convolutional
Neural Networks
(CNN), ReLU
activation, max pooling,
and the 'adam' optimizer
for deep learning on
MRI images.
Brain Tumor
Classification
Using
Convolutional
Neural Network
[14]
The paper introduced a
CNN for the automatic
classification of
common brain tumor
types using MRI
images, achieving a top
validation accuracy of
84.19% with the chosen
CNN architecture.
The CNN approach
offers efficient and
accurate brain tumor
classification without
the need for time
consuming region based
preprocessing,
simplifying the
diagnostic workflow for
medical professionals.
Includes React for the
front-end, Node.js for
the back end, MongoDB
for the database, and
Docker for
containerization.

Article No. 11
PP 01-13
11
10
Brain Tumor
Detection Using
Deep Neural
Network and
Machine Learning
Algorithm
[15]
The study demonstrates
that the Softmax
classifier within the
CNN achieved an
accuracy of 98.67% for
image categorization,
and the proposed
approach, which
combines feature
extraction and CNN,
improved accuracy to
99.12% on test data.
High accuracy achieved
in brain tumor detection,
enabling early diagnosis
and improved patient
care
Requirement for large
datasets and substantial
computational
resources, which may
limit its practical
application in some
healthcare settings.

Article No. 11
PP 01-13
12
10
IV. CONCLUSION
In conclusion, NesuroSentinel Prodigy innovative Brain Tumor Detection is designed to empower healthcare
professionals with a cutting-edge, automated solution for the early detection and precise classification of brain
tumors. By harnessing the capabilities of advanced artificial intelligence and machine learning, NesuroSentinel
Prodigy significantly streamlines the diagnostic process, accelerating the crucial intervention and treatment
planning phase. The project's adaptability and user-friendly interface make it a valuable asset for medical
practitioners and institutions, delivering faster and more accurate diagnoses. Beyond its efficiency gains, the
system enhances the quality of patient care, equipping healthcare providers with the tools to make informed
decisions and ultimately improve patient outcomes. NesuroSentinel Prodigy represents a significant stride
forward in the field of medical image analysis, offering a promising path for future research and progress in the
vital domain of brain tumor detection.
REFERENCES
[1] Tariq Sadad, Amjad Rehman, Asim Muni,r Tanzila Saba, Usman Tariq, Noor Ayesha, Rashid Abbasi “Brain tumor
detection and multi-classification using advanced deep learning techniques” 2020, National Institutes of Health.
[2] Pallavi Tiwari, Bhaskar Pant, Mahmoud M. Elarabawy, Mohammed Abd-Elnaby, Noor Mohd, Gaurav Dhiman and
Subhash Sharma “CNN Based Multiclass Brain Tumor Detection Using Medical Imaging” 2022, Hindawi.
[3] Tanzila Saba, Ahmed Sameh Mohamed, Mohammad El-Affendi, Javeria Amin, Muhammad Sharif “Brain tumor
detection using fusion of hand crafted and deep learning features” 2020, ELSEVIER.
[4] Md Khairul Islam, Md Shahin Ali, Md Sipon Miah, Md Mahbubur Rahman, Md Shahariar Alam, Mohammad Amzad
Hossain “Brain tumor detection in MR image using superpixels, principal component analysis and template-based K-
means clustering algorithm” 2021, ELSEVIER.
[5] Chirodip Lodh Choudhury, Chandrakanta Mahanty, Raghvendra Kumar “Brain Tumor Detection and Classification
Using Convolutional Neural Network and Deep Neural Network” 2020, IEEE.
[6] Sarmad Maqsood, Robertas Damaševi cius and Rytis Maskeliunas “Multi-Modal Brain Tumor Detection Using Deep
Neural Network and Multiclass SVM” 2022, MEDICINA.
[7] Mesut Toğaçar, Burhan Ergen, Zafer Cömert “BrainMRNet: Brain Tumor Detection using Magnetic Resonance Images
with a Novel Convolutional Neural Network Model” 2019, ELSEVIER.
[8] Shtwai Alsubai, Habib Ullah Khan, Abdullah Alqahtani, Mohemmed Sha, Sidra Abbas and Uzma Ghulam Mohammad
“Ensemble Deep Learning dor Brain Tumor Detection.” 2022, FRONTIERS.
[9] Wadhah Ayadi, Wajdi Elhamzi, Imen Charf, Mohamed Atri “Deep CNN for Brain Tumor Classifcation” 2021,
SPRINGER.
[10] Neelum Noreen, Sellappan Palaniappan, Abdul Qayyum, Iftikhar Ahmad, Muhammad Imran, and Muhammad Shoaib
“A Deep Learning Model Based on Concatenation Approach for the Diagnosis of Brain Tumor” 2020, IEEE.
[11] Emrah Irmak “Multi-Classification of Brain Tumor MRI Images Using Deep Convolutional Neural Network with Fully
Optimized Framework” 2020, SPRINGER.
[12] Francisco Javier Díaz-Pernas, Mario Martínez-Zarzuela, Míriam Antón-Rodríguez and David González-Ortega “A Deep
Learning Approach for Brain Tumor Classification and Segmentation Using a Multiscale Convolutional Neural
Network” 2021, MDPI.
[13] Nyoman Abiwinanda, Muhammad Hanif, S. Tafwida Hesaputra, Astri Handayani, and Tati Rajab Mengko “Brain
Tumor Classification Using Convolutional Neural Network” 2021, SPRINGER.
[14] Miss Krishna Pathak, Mr. Mahekkumar Pavthawala, Miss Nirali Patel, Mr. Dastagir Malek, Prof. Vandana Shah, Prof.
Bhaumik Vaidya “Classification of Brain Tumor Using Convolutional Neural Network” 2019, IEEE.
[15] Masoumeh Siar, Mohammad Teshnehlab “Brain Tumor Detection Using Deep Neural Network and Machine Learning

Article No. 11
PP 01-13
13
10
Algorithm” 2019, ResearchGate.
[16] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9468505/, last accessed on: 15/10/2023
[17] Wang Y, Wang L, Wang H, Li P. End-to-end image super-resolution via deep and shallow convolutional networks.
IEEE Access. 2019;7:31959–7.

NEUROSENTINEL PRODIGY (Brain Tumor) using CNN

More Related Content

Similar to NEUROSENTINEL PRODIGY (Brain Tumor) using CNN (20)

More from vivatechijri (20)

Recently uploaded (20)

NEUROSENTINEL PRODIGY (Brain Tumor) using CNN