Design of Transfer Learning based Deep CNN Paradigm for Brain
Tumor Classification
NEHA BHARDWAJ, MEENAKSHI SOOD, SS GILL
ECE Department, NITTTR, Chandigarh,
Panjab University,
INDIA
Abstract: - Brain tumor is a fatal illness causing worldwide fatalities. The existing neuroimaging modalities to
detect brain tumors are invasive and are observer-biased. Automatic CAD frameworks using sophisticated AI
techniques lessen human intervention and can effectively handle large amounts of data. Automatic CAD
frameworks using Machine learning techniques require the use of time-consuming and error-prone manual
feature extraction procedures. Deep learning techniques involve automatic feature extraction; hence,
appreciable classification results are attained quickly. However, training DL models from scratch takes a
significant investment of time, money, and large datasets, which are difficult to attain in the medical domain.
Therefore, the trade-off is utilizing the well exhaustively learned models like VGG16, VGG19, AlexNet, etc. to
design a novel framework for the classification of brain tumors. The paper aims to develop a CNN-based deep
learning framework by fine-tuning the pre-trained VGG16 architecture via transfer learning for brain tumor
detection. The designed framework employing the transfer-learning technique gives better results with less data
in less time. The brain tumor binary classification using brain MR images using transfer learning achieved an
appreciable accuracy of 97%. The training and validation accuracy obtained was 100% and 97%, respectively,
with 30 epochs. The loss for classification was as low as 0.0059% and the run time of 32ms/step time, much
less than the existing models.
Key-Words: - Convolutional Neural Networks, Deep Learning, Computer Aided Diagnosis, Classification,
Hyperparameter tuning, Magnetic Resonance Imaging.
Received: August 11, 2023. Revised: December 27, 2023. Accepted: February 5, 2024. Published: April 4, 2024.
1 Introduction
Diseases have stumbled and been vanquished over
the past decades due to human knowledge and
biomedical advancement, but cancer, due to its
unstable nature, remains a burden to mankind. Brain
tumor malignancy is a severe as well as rapidly
progressing disorder. The brain is an important and
complicated organ in the human body, comprising
nerve cells and tissues that control the majority of
the body's operations.
As per the research report published in the
International Association of Cancer Registries
(IARC), in June 2023, approximately 28000 cases
of brain tumor are reported in India each year, with
approximately 24000 fatalities. According to
research published in Business Insider by doctors,
this lethal condition affects 20% of the young Indian
population, [1]. One of the most critical aspects of
saving a person's life is the quick diagnosis and
prognosis of this hazardous disease.
There exist many neuroimaging modalities like
MRI, CT Scan, and PET Scan to detect brain
tumors, but decoding the type and grade requires
biopsy measures, which are prone to human
subjectivity. Nowadays, advancements in
Computer-Assisted Diagnosis (CAD), and AI
methods enable radiologists to identify brain tumors
more accurately. Artificial Intelligence (AI) tools
are now widely used in biomedical exploration and
the development of robust diagnostic systems for a
variety of diseases due to their success in prediction
particularly in clinical analysis to characterize brain
tumors. Outmoded machine learning methods for
disease characterization require manual feature
extraction, which brings human-biased outcomes.
Deep learning approaches, on the other hand, can be
developed in such a way that no handmade feature
extraction is required while producing correct
classification results.
Machine learning techniques use feature
abstraction methods, such as thresholding-based,
clustering-based, contour-based, and texture-based,
to segregate the tumor from normal anatomical
surroundings . However, Deep Learning techniques
tackles this issue by using automatic feature
extraction techniques through the use of
Convolutional Neural Networks (CNN), Long Short
Term Memory Networks (LSTM), and Recurrent
WSEAS TRANSACTIONS on BIOLOGY and BIOMEDICINE
DOI: 10.37394/23208.2024.21.17
Neha Bhardwaj, Meenakshi Sood, Ss Gill
E-ISSN: 2224-2902
162
Volume 21, 2024
Neural Networks (RNN) deep learning models.
Training DL models from scratch involves time and
resources and requires huge datasets, which are not
always easily available. In machine learning, it is
possible to lose significant knowledge from the
actual data by employing manual feature extraction.
Therefore, the trade-off is a transfer learning
approach, which enables using a pre-trained
classification model and fine-tuning it on a new
relatable classification problem. In transfer learning,
a model already trained with other enormous
databases linked to a different field is used for
classification. Transfer learning saves resources and
time as the model is not trained from scratch. This
knowledge enables the model to attain appreciable
accuracy on a smaller database.
This paper proposed a deep CNN-based
framework by fine-tuning the pre-existing model,
VGG16, to classify brain tumor MR images into
healthy and tumor images by fine-tuning the last
layers to 2 on a publically available benchmark
dataset. The resizing of the image dataset is done to
match the image dimension requirement of VGG16,
and pre-processing of the acquired database is done
to achieve appreciable classification accuracy with
low errors and in less time.
The paper is organized as follows: Section 2
provides the related literature survey carried out by
researchers in a similar field. Section 3 presents the
methodology of the process, starting from data
acquisition to fine-tuning and, finally, binary
classification of the brain tumor. Section 4
represents the results and discussions, and Section 5
concludes the paper.
2 Background
A brain tumor identification and classification
network using Recurrent Convolutional Neural
Network was developed and a classification
accuracy of 95.17% was obtained on the Kaggle
dataset, [2]. A survey was conducted on the
available brain tumor detection and its grading
techniques and the importance of timely diagnosis
of a tumor for saving a person’s life as tumor
changes shape and size quickly was concluded, [3].
A transfer learning technique was developed for
multi-classification of brain tumors into three types
using CNN, and classification was done using
classifiers like Support Vector Machine, and K-
Nearest Neighbour on the Figshare database, [4],
[5], [6].
A multiclassification framework using transfer
learning on VGG16 was developed and achieved a
classification accuracy of 97.80% on a publicly
available database of 3064 brain MRI images, [7]. A
deep learning-based pre-trained classification
models for binary classification of brain tumors was
developed, by fine-tuning pretrained AlexNet, and
an accuracy of 99% was achieved on The Cancer
Imaging Archive (TCIA) Public Access repository
containing 696 MR images, [8]. Two different
techniques were developed: the first, for assessing
cancer grade directly from imaging data, obtained
an accuracy of 89.5%, while the second, for
predicting grade from tumor ROI, obtained an
accuracy of 92.98%, [9]. A pre-trained model
Google Net was finetuned for multiclassification of
brain MR images using the public data repository
Figshare consisting of 3064 brain MR images and
an accuracy of 98% was achieved, [10]. A Neural
Net-based model for multi-classification of three
types of brain tumors using Keras was developed
and an accuracy of 95% on the Figshare brain MR
image dataset consisting of 3064 brain MR images
was obtained, [11].
A deep-learning CNN framework for brain
tumor detection using Keras tensorboard was
developed and classification accuracy of 99.40% on
the publically available BraTs 2020 dataset of 3064
brain MR images was obtained, [12]. The multi-
classification accuracy of 95% on the Figshare
dataset of 3,064 brain MRI databases using the pre-
trained model ResNet 50 was achieved. and a
comparative analysis with other cutting-edge
models, such as DenseNet and Mobilenet, was
performed by the authors, [13]. Two CNN
frameworks for brain tumor binary and multi-
classification using the Kaggle database were
developed, with 94% and 89% accuracy,
respectively, [14]. Image pre-processing was done
on a Kaggle dataset of brain tumor MRI images
using a grey-level co-occurrence matrix for feature
extraction and classification accuracy of 95.17%
was achieved. The study comprised of a benchmark
database of 3264 brain MRI images, [15]. The
Multimodal MRI scans were utilized to demonstrate
genomic subtyping of Glioma in the brain with an
accuracy of 82.35%. The MRI image dataset
BR35H: Brain Tumour Detection 2020 (BR35H)
was used for the same purpose, [16]. The literature
revealed that brain tumor classification and the
multi-classification of the brain tumor has been
achieved by researchers with appreciable
classification accuracy by either developing their
frameworks or by utilizing the already existing
state-of-the-art classification models like VGG16,
VGG19, AlexNet, ResNet, etc. that gives better
accuracy results in less time with limited dataset.
WSEAS TRANSACTIONS on BIOLOGY and BIOMEDICINE
DOI: 10.37394/23208.2024.21.17
Neha Bhardwaj, Meenakshi Sood, Ss Gill
E-ISSN: 2224-2902
163
Volume 21, 2024
3 Methodology
The design of the framework starts from data
acquisition to pre-processing and fine-tuning of pre-
trained VGG16 for brain tumor classification.
3.1 Data Acquisition and Data Pre-
Processing
The database is taken from the official data port
Kaggle, BR35H-Brain Tumor Detection 2020, a
Google benchmark data repository for data scientists
and machine learning practitioners, [17]. The dataset
is balanced and consists of 3000 labelled brain MRI
scans, 1500 tumors and 1500 non-tumor.
The acquired dataset is subjected to various
preprocessing techniques like resizing the dataset to
224*224 as this size image set is fed to pre-trained
VGG16. To ensure uniformity, the entire dataset is
scaled to the same value before being fed into the
neural net. The entire dataset is randomly shuffled
before being fed to the classification framework to
prevent the network from targeting specific images
every time.
The proposed CNN Model is implemented
using Tensorflow 2.6, Google’s machine learning
platform, and i5 processor with 64-bit operating
system and 16 GB RAM. The acquired database is
split into training and validation image bases. A
validation split of 0.3 or 0.2 is frequently utilized for
the split. It is a simple procedure for observing the
enactment of predictive deep or machine learning
models. The training image database is used to fit
the deep net or machine-learning model, whereas
the testing dataset is used to evaluate the model. In
this brain tumor categorization paradigm, a data
split of 0.3 has been used, [18].
The block diagram for brain tumor
identification is presented in Figure 1.
Fig. 1: Block Diagram of brain tumor detection using VGG16 by the process of transfer learning
Brain MR Image
dataset
(BraTs 2020)
Data Pre-
processing
Data Resizing
Data Normalization
Data Shuffling
Data Division
Validation Split =0.3
Training 70%
Testing 30%
Tumor MRI
Healthy MRI
Fine-tuned VGG16
Binary Classification
WSEAS TRANSACTIONS on BIOLOGY and BIOMEDICINE
DOI: 10.37394/23208.2024.21.17
Neha Bhardwaj, Meenakshi Sood, Ss Gill
E-ISSN: 2224-2902
164
Volume 21, 2024
3.2 Transfer Learning
The transfer learning method uses pre-trained
classification models for learning new features
to use the already learned features to solve one
problem as a starting point for solving other
problems with a new dataset.
VGG16 is a 16-layer deep Convolutional
Neural Network developed by University of
Oxford, [19]. When compared to other evolved
comprehensives, it stands out for its simplicity.
The acceptable input data size for VGG 16 is
224*224 pixels. It has 13 Convolutional layers,
5 max pooling layers, and 3 dense layers. The
filter size varies from 64 at the start to 512 at
end. The output layer has 1000 neurons
accounting for 1000 classes of the ImageNet
dataset, VGG16 was trained upon, [20], [21].
3.3 Proposed Fine-tuned Classification
Paradigm
In this research work, the pre-trained model
VGG16 has been fine-tuned by changing the
last layers according to the problem
undertaken, and neurons in the output layer
have been changed to two to detect the brain
tumor (binary classification). The activation
function used is Softmax.
Transfer learning helps in fast training of the
model, as the model is not trained from scratch,
hence appreciable results are attained in less
time, [22], [23]. The computational cost and
time are also saved and model gives good
results on fewer datasets. The transfer learning,
sometimes may add biases from the previous
models it was trained and also may sometimes
add the problem of overfitting. Dropout layers
are added to prevent the problem of
overlearning and underlearning. The fine-
tuning strategy in transfer learning allows
tweaking hyperparameters like learning rate
and regularisation to optimize the model as per
the requirement, [24], [25].
4 Results and Discussions
4.1 Evaluation Metrics
To inspect the performance of finetuned
VGG16 architecture, various evaluation
parameters like Accuracy, Precision, Recall,
and F1 Score are observed.
These are indications of how accurate the
prediction of the designed model is. The score
indicates better model performance.
(1)
(2)
(3)
(4)
(5)
where: TP, TN, FP, and FN are True Positive,
True Negative, False Positive, and False
Negative respectively, [26], [27].
TP: when the model correctly interprets a
normal brain image as a healthy image.
TN: when the model correctly interprets brain
tumor image as tumor image.
FP: when the model incorrectly interprets the
normal brain image.
FN: when the model incorrectly interprets the
tumor brain image.
4.2 Classification Results
The empirical exhaustive study was done by
varying the number of epochs from 10 to 50.
The overall accuracy for binary classification is
92% for 10 and 97% for 30 epochs. After
increasing the epochs to 50, the classification
accuracy remains constant, which shows
optimum learning in 30 epochs. Further, the
classification accuracy was justified by other
related metrics like F1 Score, Recall, and
Precision, [28].
The classification reports for 10 and 30
epochs are shown in Table 1 and Table 2,
respectively.
Table 1. Classification Report: (10 epochs)
Precision
Recall
F1
Score
Support
0 (Non tumor)
0.90
0.96
0.93
468
1 (Tumor)
0.96
0.88
0.92
432
Accuracy
0.92
900
Macro
Average
0.93
0.92
0.92
900
Weighted
Average
0.93
0.92
0.92
WSEAS TRANSACTIONS on BIOLOGY and BIOMEDICINE
DOI: 10.37394/23208.2024.21.17
Neha Bhardwaj, Meenakshi Sood, Ss Gill
E-ISSN: 2224-2902
165
Volume 21, 2024
Table 2. Classification Report: (30 epochs)
Precision
Recall
F1
Score
Support
0 (Non tumor)
0.98
0.96
0.97
460
1 (Tumor)
0.96
0.98
0.97
440
Accuracy
0.97
900
Macro
Average
0.97
0.97
0.97
900
Weighted
Average
0.97
0.97
0.97
The training and validation accuracy for
binary classification with 10 epochs is 99% and
92.44%, respectively. The loss and run time
was also low, accounting for 0.060 % loss and
33ms/step time. The training and validation
accuracy was 100% and 96.78%, respectively,
with 30 epochs. The loss was also low,
accounting for 0.0059% and a low run time of
32ms/step time. The accuracy graph attained a
constant value even after increasing the epochs
to 50, showing optimum learning achieved by
the network, as shown in Figure 3(c).
The Confusion Matrix is an important
performance-measuring statistic for the deep
learning model, [29], [30]. It is used to
summarise the predicted and actual values of
the developed framework. The confusion
matrix of the developed model for both 10 and
30 epochs is shown in Figure 2.
The developed model was tested on 60 MR
images of tumor and non-tumor images. The
developed framework predicted the tumor and
healthy images correctly. The testing results of
tumor and healthy images by the developed
framework are shown in Figure 4, respectively.
(a)
(b)
Fig. 2: Confusion Matrix of developed Model
(a) 10 epochs, (b) 30 epochs
(a)
(b)
(c)
Fig. 3: ROC Curve of the developed model (a)
10 epochs (b) 30 epochs (c) 50 epochs
WSEAS TRANSACTIONS on BIOLOGY and BIOMEDICINE
DOI: 10.37394/23208.2024.21.17
Neha Bhardwaj, Meenakshi Sood, Ss Gill
E-ISSN: 2224-2902
166
Volume 21, 2024
Fig. 4: Testing of developed model on (a)
tumor and (b) Non tumor MR image
4.3 Comparative Analysis
The comparative analysis of the fine-tuned
VGG 16 for brain tumor classification has been
done by the classification frameworks
developed by the authors over the time for
brain tumor detection using the same dataset
used in this research study as shown in Table 3.
The percentage improvement in the accuracy
for the developed model with the frameworks
developed by the authors with the same
database is shown in Figure 5.
Table 3. Comparative Analysis
Mod
el
Nam
e
Methodol
ogy
Accura
cy for
Model
develop
ed by
authors
Classificat
ion
accuracy
achieved
by
developed
Model
Percentag
e
improvem
ent in the
accuracy
M1
T.Tazin et
al. [21]
92%
97%
5.15%
M2
Agas Eko
et al. [22]
97%
97%
-
M3
F. Özyurt
et al. [23]
95.62%
97%
1.42%
M4
Pereira et
al. [9]
89.50%
97%
7.73%
M5
O.zkaraca
et al. [14]
94%
97%
3.09%
Fig. 5: Percentage Improvement Accuracy
Graph
5 Conclusion and Future Work
Brain tumors are lethal, and manual discovery
takes time and medical expertise. The large
volume of MR data required for tumor
diagnosis and type necessitates using automatic
diagnostic approaches. The paper presented a
fine-tuned VGG 16 architecture using the
transfer learning process for automatic brain
tumor uncovering in human brain MR images.
The results demonstrated the effective
automatic brain tumor binary classification
without human intervention. The results of the
F1 Score, precision, Recall, and high training
and testing accuracy with a low test error of
0.0059% and low runtime of 32ms/step time.
The overall accuracy obtained was 92% for 10
epochs and 97% for 30 epochs and it remains
constant after increasing the number of epochs
to 50.
In the future, a novel CNN-based classification
model will be developed for brain tumor
severity grading which will further add to the
development of robust CAD systems for multi-
classification. The results obtained will be
compared with the existing classification
models like VGG16, VGG19, AlexNet,
ResNet, etc.
Acknowledgment:
The authors thank NITTTR, Chandigarh, and
AICTE for carrying out this research.
References:
[1] “Brain Tumor: Symptoms, Signs &
Causes”, [Online].
https://my.clevelandclinic.org/health/dise
ases/6149-brain-cancer-brain-tumor
(Accessed Date: November 16, 2023).
0,00%
5,00%
10,00%
M1 M2 M3 M4 M5
Percentage improvement in Accuracy
% improvement
WSEAS TRANSACTIONS on BIOLOGY and BIOMEDICINE
DOI: 10.37394/23208.2024.21.17
Neha Bhardwaj, Meenakshi Sood, Ss Gill
E-ISSN: 2224-2902
167
Volume 21, 2024
[2] H. Alsaif, R. Guesmi, and Alshammari,
“A Novel Data Augmentation-Based
Brain Tumor Detection Using
Convolutional Neural Network,” Appl.
Sci., vol. 12, no. 8, 2022, doi:
10.3390/app12083773.
[3] N. Bhardwaj, M. Sood, and S. S. Gill,
“Artificial Intelligence-Empowered 3D
Bioprinting,” AI Big Data-Based Eng.
Appl. from Secur. Perspect., pp. 1–20,
Jun. 2023, doi: 10.1201/9781003230113-
1.
[4] J. O. Healthcare Engineering, “Retracted:
Brain Tumor Detection and Classification
by MRI Using Biologically Inspired
Orthogonal Wavelet Transform and Deep
Learning Techniques,” J. Healthc. Eng.,
vol. 2023, p. 9845732, 2023, doi:
10.1155/2023/9845732.
[5] H. Alsaif, R. Guesmi, and Alshammari,
“A Novel Data Augmentation-Based
Brain Tumor Detection Using
Convolutional Neural Network,” Appl.
Sci., vol. 12, no. 8, 2022, doi:
10.3390/app12083773.
[6] H. C. Shin, H. R. Roth, and Gao, “Deep
Convolutional Neural Networks for
Computer-Aided Detection: CNN
Architectures, Dataset Characteristics and
Transfer Learning,” IEEE Trans. Med.
Imaging, vol. 35, no. 5, pp. 1285–1298,
2016, doi: 10.1109/TMI.2016.2528162.
[7] M. S. I. Khan and Rahman, “Accurate
brain tumor detection using deep
convolutional neural network,” Comput.
Struct. Biotechnol. J., vol. 20, pp. 4733–
4745, 2022, doi:
10.1016/j.csbj.2022.08.039.
[8] R. Mehrotra, M. A. Ansari, and Agrawal,
“A Transfer Learning approach for AI-
based classification of brain tumors,”
Mach. Learn. with Appl., vol. 2, no.
October, p. 100003, 2020, doi:
10.1016/j.mlwa.2020.100003.
[9] F. Pereira, B. Lou, and Pritchett,
“Toward a universal decoder of linguistic
meaning from brain activation,” Nat.
Commun., vol. 9, no. 1, 2018, doi:
10.1038/s41467-018-03068-4.
[10] S. Deepak and P. M. Ameer, “Brain
tumor classification using deep CNN
features via transfer learning,” Comput.
Biol. Med., vol. 111, no. March, p.
103345, 2019, doi:
10.1016/j.compbiomed.2019.103345.
[11] H. Ucuzal, Ş. YAŞAR and C. Çolak,
"Classification of brain tumor types by
deep learning with convolutional neural
network on magnetic resonance images
using a developed web-based
interface," 2019 3rd International
Symposium on Multidisciplinary Studies
and Innovative Technologies (ISMSIT),
Ankara, Turkey, 2019, pp. 1-5, doi:
10.1109/ISMSIT.2019.8932761.
[12] N. Bhardwaj, M. Sood and S. Gill, "Deep
Learning Framework using CNN for
Brain Tumor Classification," 2022 5th
International Conference on Multimedia,
Signal Processing and Communication
Technologies (IMPACT), Aligarh, India,
2022, pp. 1-5, doi:
10.1109/IMPACT55510.2022.10029043.
[13] T. Sadad and Rehman, “Brain tumor
detection and multi-classification using
advanced deep learning techniques,”
Microsc. Res. Tech., vol. 84, no. 6, pp.
1296–1308, 2021, doi:
10.1002/jemt.23688.
[14] O. Özkaraca and Bağrıaçık, “Multiple
Brain Tumor Classification with Dense
CNN Architecture Using Brain MRI
Images,” Life, vol. 13, no. 2, 2023, doi:
10.3390/life13020349.
[15] R. Vankdothu and Hameed, “Brain tumor
MRI images identification and
classification based on the recurrent
convolutional neural network,” Meas.
Sensors, vol. 24, no. August, p. 100412,
2022, doi:
10.1016/j.measen.2022.100412.
[16] P. Tupe-Waghmare, P. Malpure, and
Kotecha, “Comprehensive Genomic
Subtyping of Glioma Using Semi-
Supervised Multi-Task Deep Learning on
Multimodal MRI,” IEEE Access, vol. 9,
pp. 167900–167910, 2021, doi:
10.1109/ACCESS.2021.3136293.
[17] “Br35H :: Brain Tumor Detection 2020 |
Kaggle”, [Online].
https://www.kaggle.com/datasets/ahmedh
amada0/brain-tumor-detection (Accessed
Date: August 3, 2022).
[18] S. Srinivasan and P. S. M. Bai, “Grade
Classification of Tumors from Brain
Magnetic Resonance Images Using a
Deep Learning Technique,” Diagnostics,
vol. 13, no. 6, 2023, doi:
10.3390/diagnostics13061153.
[19] S. Liu and W. Deng, "Very deep
WSEAS TRANSACTIONS on BIOLOGY and BIOMEDICINE
DOI: 10.37394/23208.2024.21.17
Neha Bhardwaj, Meenakshi Sood, Ss Gill
E-ISSN: 2224-2902
168
Volume 21, 2024
convolutional neural network based
image classification using small training
sample size," 2015 3rd IAPR Asian
Conference on Pattern Recognition
(ACPR), Kuala Lumpur, Malaysia, 2015,
pp. 730-734, doi:
10.1109/ACPR.2015.7486599.
[20] “6 Types of Classifiers in Machine
Learning | Analytics Steps”, [Online].
https://www.analyticssteps.com/blogs/typ
es-classifiers-machine-learning
(Accessed Date: May 26, 2022).
[21] T. Tazin, S. Sarker, P. Gupta, and Gupta,
“A Robust and Novel Approach for Brain
Tumor Classification Using
Convolutional Neural Network,”
Comput. Intell. Neurosci., vol. 2021,
2021, doi: 10.1155/2021/2392395.
[22] A. S. M. Shafi, M. B. Rahman, and
Anwar, “Classification of brain tumors
and auto-immune disease using ensemble
learning,” Informatics Med. Unlocked,
vol. 24, p. 100608, 2021, doi:
10.1016/j.imu.2021.100608.
[23] F. Özyurt, E. Sert, and Avci, “Brain tumor
detection based on Convolutional Neural
Network with neutrosophic expert
maximum fuzzy sure entropy,” Meas. J.
Int. Meas. Confed., vol. 147, 2019, doi:
10.1016/j.measurement.2019.07.058.
[24] H. U. Xiuqiong, “A New Optimal Power
Flow Model Considering the Active
Power Constraints of Transmission
Interfaces,” WSEAS Transactions on
Circuits and Systems, vol. 22, pp. 16–22,
2023,
https://doi.org/10.37394/23201.2023.22.3
[25] E. S. Ali and S. M. Abd Elazim, “Power
System Stability Enhancement Using
Grasshopper Optimization Approach and
PSSs,” WSEAS Transactions on Power
Systems, vol. 18, no. 1, pp. 135–140,
2023,
https://doi.org/10.37394/232016.2023.18.
14.
[26] A. K. Daud and S. Khader, “Closed Loop
Modified SEPIC Converter for
Photovoltaic System,” WSEAS
Transactions on Circuits and Systems,
vol. 21, no. D, pp. 161–167, 2022,
https://doi.org/10.37394/23201.2022.21.1
7.
[27] A. Zemliak, “On the Structure of a Quasi-
Optimal Algorithm for Circuit
Designing,” WSEAS Transactions on
Circuits and Systems, vol. 21, pp. 168–
175, 2022,
https://doi.org/10.37394/23201.2022.21.1
8.
[28] H. A. Shah, F. Saeed, and Yun, “A
Robust Approach for Brain Tumor
Detection in Magnetic Resonance Images
Using Finetuned EfficientNet,” IEEE
Access, vol. 10, pp. 65426–65438, 2022.
[29] H. H. Sultan, N. M. Salem and W. Al-
Atabany, "Multi-Classification of Brain
Tumor Images Using Deep Neural
Network," in IEEE Access, vol. 7, pp.
69215-69225, 2019, doi:
10.1109/ACCESS.2019.2919122.
[30] Bhardwaj, N., Sood, M., Gill, S.S.
(2024). Data Pre-processing Techniques
for Brain Tumor Classification. In:
Mehta, G., Wickramasinghe, N., Kakkar,
D. (eds) Innovations in VLSI, Signal
Processing and Computational
Technologies. WREC 2023. Lecture
Notes in Electrical Engineering, vol
1095. Springer, Singapore.
https://doi.org/10.1007/978-981-99-7077-
3_20.
Contribution of Individual Authors to the
Creation of a Scientific Article
The authors participated equally in the current
study at all stages, from problem
conceptualization to results and solution.
Sources of Funding for Research Presented
in a Scientific Article or Scientific Article
Itself
No funding was received for conducting this
study.
Conflict of Interest
The authors have no conflicts of interest to
declare.
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the
Creative Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/de
ed.en_US
WSEAS TRANSACTIONS on BIOLOGY and BIOMEDICINE
DOI: 10.37394/23208.2024.21.17
Neha Bhardwaj, Meenakshi Sood, Ss Gill
E-ISSN: 2224-2902
169
Volume 21, 2024
