Skip to main content <#maincontent>
We will keep fighting for all libraries - stand with us!
Internet Archive logo A line drawing of the Internet Archive
headquarters building façade.
Search icon An illustration of a magnifying glass.
Search icon An illustration of a magnifying glass.
Upload icon An illustration of a horizontal line over an up pointing
arrow. Upload
User icon An illustration of a person's head and chest. Sign up
| Log in
Web icon An illustration of a computer application window
Wayback Machine
Texts icon An illustration of an open book.
Books
Video icon An illustration of two cells of a film strip.
Video
Audio icon An illustration of an audio speaker.
Audio
Software icon An illustration of a 3.5" floppy disk.
Software
Images icon An illustration of two photographs.
Images
Donate icon An illustration of a heart shape
Donate
Ellipses icon An illustration of text ellipses.
More
Hamburger icon An icon used to represent a menu that can be toggled by
interacting with this icon.
Internet Archive Audio
Live Music Archive Librivox Free
Audio
Featured
* All Audio
* This Just In
* Grateful Dead
* Netlabels
* Old Time Radio
* 78 RPMs and Cylinder Recordings
Top
* Audio Books & Poetry
* Computers, Technology and Science
* Music, Arts & Culture
* News & Public Affairs
* Spirituality & Religion
* Podcasts
* Radio News Archive
Images
Metropolitan Museum
Cleveland
Museum of Art
Featured
* All Images
* This Just In
* Flickr Commons
* Occupy Wall Street Flickr
* Cover Art
* USGS Maps
Top
* NASA Images
* Solar System Collection
* Ames Research Center
Software
Internet Arcade Console
Living Room
Featured
* All Software
* This Just In
* Old School Emulation
* MS-DOS Games
* Historical Software
* Classic PC Games
* Software Library
Top
* Kodi Archive and Support File
* Vintage Software
* APK
* MS-DOS
* CD-ROM Software
* CD-ROM Software Library
* Software Sites
* Tucows Software Library
* Shareware CD-ROMs
* Software Capsules Compilation
* CD-ROM Images
* ZX Spectrum
* DOOM Level CD
Books
Books to Borrow Open Library
Featured
* All Books
* All Texts
* This Just In
* Smithsonian Libraries
* FEDLINK (US)
* Genealogy
* Lincoln Collection
Top
* American Libraries
* Canadian Libraries
* Universal Library
* Project Gutenberg
* Children's Library
* Biodiversity Heritage Library
* Books by Language
* Additional Collections
Video
TV News Understanding 9/11
Featured
* All Video
* This Just In
* Prelinger Archives
* Democracy Now!
* Occupy Wall Street
* TV NSA Clip Library
Top
* Animation & Cartoons
* Arts & Music
* Computers & Technology
* Cultural & Academic Films
* Ephemeral Films
* Movies
* News & Public Affairs
* Spirituality & Religion
* Sports Videos
* Television
* Videogame Videos
* Vlogs
* Youth Media
Search the history of over 835 billion web pages
on the Internet.
Search the Wayback Machine
Search icon An illustration of a magnifying glass.
Mobile Apps
* Wayback Machine (iOS)
* Wayback Machine (Android)
Browser Extensions
* Chrome
* Firefox
* Safari
* Edge
Archive-It Subscription
* Explore the Collections
* Learn More
* Build Collections
Save Page Now
Capture a web page as it appears now for use as a trusted citation in
the future.
Please enter a valid web address
* About
* Blog
* Projects
* Help
* Donate
* Contact
* Jobs
* Volunteer
* People
* Sign up for free
* Log in
Search metadata
Search text contents
Search TV news captions
Search radio transcripts
Search archived web sites
Advanced Search
* About
* Blog
* Projects
* Help
* Donate Donate icon An illustration of a heart shape
* Contact
* Jobs
* Volunteer
* People
Full text of "Monkeypox Detection Using Hyper-Parameter Tuned Based
Transferable CNN Model
"
See other formats
Int. J. Exp. Res. Rev., Vol. 33: 18-29 (2023)
Open Access
| 2455-4855
08
DOI: https://doi.org/10.52756/ijerr.2023.v33spl.003
Original Article
Peer Reviewed
International Journal of Experimental Research and Review (IJERR) il
© Copyright by International Academic Publishing House (IAPH)
ISSN: 2455-4855 (Online)
www. iaph.in
al 48
Monkeypox Detection Using Hyper-Parameter Tuned Based Transferable CNN Model
V. Gokula Krishnan", B. S. Liya*, S. Venkata Lakshmi, K. Sathyamoorthy*
and Sangeetha Ganesan® ® Check for updates
‘Department of Computer Science and Engineering, Saveetha School of Engineering, Saveetha Institute of Medical and Technical
Sciences, Thandalam, Chennai, Tamil Nadu, India; 7Department of Computer Science and Engineering, Easwari Engineering
College, Ramapuram, Chennai, Tamil Nadu, India; *Department of AIDS, Sri Krishna College of Engineering and Technology,
Coimbatore, Tamil Nadu, India; ‘Department of Computer Science and Engineering, Panimalar Engineering College,
Poonamallee, Chennai, Tamil Nadu, India; “Department of AIDS, R M K College of Engineering and Technology, Kavaraipettai,
Tamil Nadu, India
E-mail/Orcid Id:
VGK, @gokul_kris143 @yahoo.com,® https://orcid.org/0009-0005-68 19-6729; BSL® bsliyapraveen @ gmail.com; SVL® venkatalakshmis227 @ gmail.com;
Ks,® pitsathyamoorthy @ gmail.com; SG, @ gsangeethakarthik @ gmail.com
Article History:
Received: 25" Jul., 2023
Accepted: 17" Sep., 2023
Published: 30" Sep., 2023
Abstract: The reemergence of Monkeypox, a communicable illness resulting from the
Monkeypox virus, has raised apprehensions about a potential swift global pandemic similar
to the COVID-19 epidemic as COVID-19 infections diminish globally. The prompt
emphasizes the criticality of prompt action within communities to mitigate the
development of the phenomenon. The timely identification and accurate categorization of
Monkeypox cutaneous manifestations are crucial for the successful implementation of
containment strategies. This paper presents a novel methodology for detecting Monkeypox
by utilizing a transferrable Convolutional Neural Network (CNN) model that has been
Keywords:
Chickenpox,
Convolutional Neural
Network, Modified Gear optimized utilizing hyper-parameter tuning techniques. The proposed methodology
and Steering based initiates by improving the quality of the original Monkeypox images, with a specific
Model, Monkeypox emphasis on boosting edge details to increase visual clarity. Texture qualities are obtained
Virus, Rider through an energy layer, enhancing distinctive traits. Our methodology's cornerstone is
Optimization Algorithm, utilizing the Hyper-parameter-based transferable Convolutional Neural Network (HPT-
araranisrerible TCNN), specifically designed to enhance classification accuracy.
In contrast to traditional methods, we enhance the architectural design by replacing the
HowitorciterthistNrticle: pooling layer with a configuration comprising three convolutional layers and one energy
Re ear eanerae me layer. The hyper-parameter tuning procedure is optimized by employing the Optimisation
Sathyamoorthy and Sangeetha
Ganesan (2023). Monkeypox
detection using hyper-parameter
tuned based transferable CNN
model. International Journal of
Experimental Research and
Review, 33, 18-29.
DOI: _ https://doi.org/10.52756/
ijerr.2023.v33spl.003
Introduction
The etiological agent responsible for Monkeypox,
known as the Monkeypox virus, is classified under the P
*Corresponding Author: gokul_kris143 @yahoo.com
This work is licensed under a Creative Commons Attribu-
tion-NonCommercial-NoDerivatives 4.0 International License.
Algorithm known as MGS-ROA. In order to enhance the process of model training and
validation, we have assembled the "Monkeypox Skin Lesion Dataset (MSLD)," which
consists of a collection of images depicting human skin lesions produced by Monkeypox.
The dataset in question is vital in evaluating and improving our methodology. In a
comparison analysis conducted on other deep learning models, the suggested model has
superior performance compared to other models, obtaining a notable accuracy, sensitivity,
and specificity, all reaching a value of 93.60%. The outstanding performance shown in this
study highlights the methodology's effectiveness in adequately classifying skin lesions
associated with Monkeypox. This approach shows potential for physicians and healthcare
workers since it facilitates early detection, a crucial factor in preventing the spread of
Monkeypox.
family and the orthopoxviral genus. The variola virus is
known to induce the development of smallpox within
individuals belonging to the same familial lineage
Int. J. Exp. Res. Rev., Vol. 33: 18-29 (2023)
(Samaranayake & Anil, 2022; Zumla et al., 2022). The
aetiology of bovine smallpox may be attributed to the
Cowpox virus, whereas the Vaccinia virus is employed in
developing smallpox vaccines. It is noteworthy that,
contrary to its nomenclature, the virus responsible for the
manifestation of Monkeypox has undergone evolutionary
development in rodent species. The nomenclature
"Monkeypox" (Adalja & Inglesby, 2022) was used in
1958 upon the initial discovery of the virus during two
distinct instances of outbreaks, whereby the exhibited
symptoms bore a resemblance to those employed in
scientific investigations. The discovery that the virus may
infect people was made in 1970 by researchers.
Following this, the incidence of Monkeypox has been
relatively infrequent throughout Africa. This
phenomenon has predominantly been documented in
West and Central Africa, areas distinguished by vast
tropical rainforests (Otu et al., 2022; Catala et al., 2022).
Nevertheless, during the past several years, the disease
has exhibited an increased scope, impacted a more
heterogeneous demographic and manifested in a broader
geographic expanse compared to previous occurrences.
Given the profound ramifications of the pandemic, there
has been a heightened level of attentiveness in the
surveillance of Monkeypox occurrences despite the
absence of an epidemic scale thus far (Zhang et al.,
2021).
Monkeypox is characterized by a rash that develops
over 1-5 days. The rash initially appears on the face and
subsequently spreads to other body regions. Lesions in
the vaginal region, eyes, and intraoral mucosa have been
reported in certain patients (Suganyadevi et al., 2022).
The rashes caused by this condition can seem like those
caused by chickenpox, leading to misdiagnoses. These
rashes start out as water-filled blisters but eventually heal
into crusty areas. Some people get hundreds of blisters all
over their bodies, whereas others only have a few (Ayca
et al., 2022). Lesions may join together to form
widespread rashes on the skin's surface in severe
situations. In 2-4 weeks, contingent on the harshness of
the sickness, the rashes subside, and the disease recovers.
CNNs are widely employed in academic research in
learning (Li & Du, 2021), which is revealed when we
look at the most cutting-edge technologies in image
classification. In most cases, images provide the input
data for CNN, a deep-learning model. It records the
results of several processes on the picture in order to
categorize potential future judgements. LeNet, initially
proposed by Yann LeCun in 1988 and refined until 1998
(Savas, 2022), was the first structure for a convolutional
neural network. Many industries, including NLP and
DOI: https://doi.org/10.52756/ijerr.2023.v33spl.003
biology, make use of CNN algorithms, particularly in the
realm of picture and sound processing. The best
consequences have been achieved, particularly in image
processing. The error rate was decreased to 0.23% using
CNN on the MNIST dataset.
The skin sores caused by Monkeypox are the disease's
most noticeable symptoms. In order to start therapy as
soon as possible, it is crucial to quickly distinguish skin
lesions from other lesion diseases (Saanat et al., 2022).
Mobile devices should be able to tell the difference
between Monkeypox and other illnesses that cause skin
lesions, reducing the likelihood of transmission. The end
user may determine whether or not they have Monkeypox
by taking a picture with their phone and running it
through the transfer learning-trained TFLite model (Yue
et al., 2022). Since Monkeypox has begun to spread
rapidly worldwide, we hope that our research will help
scientists swiftly and accurately categorize the impact of
this virus on skin lesions (Uysal, 2023).
In modern CAD systems, deep learning methods have
been integrated to improve the accuracy of skin lesion
identification and categorization. When it comes to
diagnosing Monkeypox, computer-aided design (CAD)
approaches based on image processing are on the rise.
The following are some of the benefits of our proposed
method for early identification of Monkeypox:
e We HPT-TCNN for
classification, with hyper-parameter tuning handled by
an MGS-ROA model.
e In the suggested HPT-TCNN architecture, we use the
energy layer (EL). By doing so, we can maintain
textural information and restrict the model's general
present an Monkeypox
capacity for learning.
Here is how the break of the paper is structured: In
Section 2, we describe the relevant literature, and in
Section 3, we offer a summary of the suggested model. In
Section 4, the trial analysis and validation are presented,
and in Section 5, the conclusion is illustrated.
Related works
Uysal et al. (2023) have created a hybrid AI system
that can identify photos of Monkeypox on the skin.
Images of skin were taken from a publicly available
picture resource. The chickenpox measles, classes make
up the multi-class structure of this dataset. The original
dataset had an uneven distribution of data across classes.
Many data augmentation and pre-processing methods
were used to correct this discrepancy. Following these
steps, state-of-the-art deep learning models, including
CSPDarkNet, RepVGG, were used to search for signs of
Monkeypox. By combining the two _best-performing
Int. J. Exp. Res. Rev., Vol. 33: 18-29 (2023)
models with the (LSTM) model, this study developed a
novel hybrid deep-learning model with improved
classification results. The built and suggested hybrid AI
system for Monkeypox detection achieved an impressive
87% test accuracy and a Cohen's kappa value of 0.8222.
Altun et al. (2023) envision using deep learning to
rapidly and precisely identify monkeypoxes from skin
lesions during the pandemic. Tools for hyperparameter
optimisation and transfer learning were made available to
support deep learning techniques. Through careful tuning
of the hyperparameters, we created a hybrid function
learning model for use within the CNN architecture.
Models, ResNETSO, and Xception were used in the
implementation. In this research, we compared and
contrasted different methods using loss and Fl-score. The
optimized hybrid MobileNetV3-s model obtained the best
results, which had an F1-score of 0.98, an average. In this
research, a bespoke CNN model was constructed using
convolutional neural networks, hyperparameter
optimization, and a hybrid function transfer learning
model, yielding impressive results. Our proposed custom
CNN model architecture demonstrates the efficiency and
effectiveness of deep learning techniques for
classification and discrimination.
Using computer vision, (Almufareh et al., 2023)
propose a smarter and more secure alternative to
conventional ways of diagnosing MPX by analyzing
photographs of skin lesions. The suggested approach
utilizes deep learning strategies to identify MPXV
positivity in skin lesions. We test our approach on two
datasets, including images and_ descriptions of
Monkeypox lesions: the (MSID). Multiple deep learning
models' performance was measured by their sensitivity to
change, specificity, and overall accuracy. Results from
using the suggested approach to identify Monkeypox
have been very encouraging, showing that it is probable
to be used on a large scale. This clever and low-cost
option may be put to good use even in underdeveloped
regions where there is a need for more laboratory
facilities.
To identify the attendance of the Monkeypox virus in
skin lesion photos, (Pramanik et al., 2023) offer an
ensemble learning-based system. First, we focus on fine-
tuning the Monkeypox dataset using one of three pre-
trained base Xception, or DenseNet169. We also use the
deep models to extrapolate probabilities that are fed into
the ensemble framework. To learn information collected
from the sum rule-based ensemble, we offer a Beta
function-based normalization scheme of probabilities to
combine the results. Using a five-fold cross-validation
configuration, the framework is extensively tested on a
DOI: https://doi.org/10.52756/ijerr.2023.v33spl.003
publicly accessible Monkeypox skin lesion dataset. The
average values for the representation’s accuracy,
precision, recall, and Fl are 93.39, 88.91, 96.78, and
92.35.
Yasmin et al. (2023) have set out to solve this issue by
using machine learning and image processing techniques
to create a model for diagnosing Monkeypox. Data
augmentation methods have been used to achieve this
goal and prevent the model from becoming overfit. Six
distinct Deep Learning (DL) models were then trained
using the pre-processed dataset using the transfer-
learning approach. We settled on the best one after
comparing each model's precision, recall, and accuracy
performance matrices. After doing fine-tuning on the
best-performing model, a new model named "PoxNet22"
was suggested. Compared to other approaches,
PoxNet22's categorization of Monkeypox is superior
since it achieves perfect results in accuracy. Clinicians
will find the findings of this study instrumental in the
classification and diagnosis of Monkeypox disease.
(2023) presented an
classification to differentiate measles. In order to model
images, researchers employed a deep learning approach
Ariansyah et al. image
based on learning. With transfer learning, a model
learned on one dataset may be applied to another. This
enabled the model to generalize insights from one data
set to another. Because deep learning is so effective for
recognizing patterns in similar photos, researchers have
proposed using it to forecast fresh data. Consequently, the
VGG-16 model achieves a respectable 83.333% accuracy
at epoch = 15.
Proposed work
Monkeypox images are fed into a DL model
optimized using the hybrid rider optimization approach
for binary classification.
Dataset
The fast spread of Monkeypox to more than 65
countries has caused public health officials to worry.
Stopping its fast development requires prompt clinical
identification. However, there needs to be more ready
access to a large sum of (PCR) tests and other
biochemical assays (Nolen et al., 2016). Computer vision
techniques might help in detecting Monkeypox from
pictures of skin lesions. However, at present, no such data
is available. Therefore, the “(MSLD)" is created by
compiling and analyzing images from various online
resources (such as websites, portals, and public case
reports). To distinguish Monkeypox patients from similar
non-Monkeypox instances, the “Lesion Dataset" was
created. Due to their resemblance to the Monkeypox rash
0
100
Int. J. Exp. Res. Rev., Vol. 33: 18-29 (2023)
200
Figure 1. Model dataset
and first-state pustules, we included lesion images of the
"Chickenpox" category to create a classifier. For
example, the Monkeypox Skin Lesion Dataset has 228
images, 102 of which are labelled as "Monkeypox" and
the residual 126 as "Others," which comprise suitcases of
non-Monkeypox. Raw data samples are exposed in
Figure 1.
Classification using DL network architecture
We the HPT-TCNN framework for
classifying cases of Monkeypox. The proposed deep
CNN considers three critical aspects of the picture: First,
introduce
certain description filters can still discover them if their
size is the same as the size filter's mask. Second, distinct
input image regions might use distinct forms or patterns.
Convolution of the full source picture is another way to
define such models. Finally, the max-pooling layer relies
heavily on down-sampled pixels, which do not alter the
original image's overall form. Figure 2 depicts the
suggested HPT-TCNN architecture for the categorization
of Monkeypox.
In the proposed HPT-TCNN, a third convolution layer
regulates the EL after two convolution layers and a
pooling layer. After the softmax layer is added. Elastic
net (EL) summarises the feature maps generated by the
DOI: https://doi.org/10.52756/ijerr.2023.v33spl.003
output of the corrected activation layer. For each feature
map, you get a number that stands for the energy
response of a filter bank. This layout reduces memory
and computational requirements and increases efficiency
while learning texture functions, EL speed and processing
time. The primary motivation for implementing this layer
is to flow. After the final pooling layer, the output of EL
is flattened and sent to the layer. Because of this link, a
new, simplified vector representing the image's contours
and textures is generated and sent to the fully linked
layer. Equation (1) delivers a size:
_ Iqg~Ip+25
Output = rr (1)
where J, and J, characterize filter size
correspondingly, S denotes the stuffing, and @ is the
stride value.
After that, 16 and 32 channels are produced by the
layers, with a kernel scope of 5 5 for the first two layers.
Using 33 kernel and 64 output channels, the convolution
layer is analyzed as a potential intermediary layer for
extracting texture attributes. From the convolution layer,
we can only parameters, which we do so by solving the
following equations:
fy =X Tp XO) ooo ceeeeeeees (2)
by = Gy FX XOX Cy ceccccccecceeceeceenens (3)
Int. J. Exp. Res. Rev., Vol. 33: 18-29 (2023)
128X1024
Benign/
Malignant
Figure 2. The framework of the projected HPT-TCNN for Monkeypox classification
Where €, symbolizes the CNN parameters, [,
characterizes the kernel extent, and ¢y denotes the
channel quantity.
The neuron output linked to the input is intended at
each convolution layer. The computation is the dot
product of the object's mass and the width of its smallest
input field. The result of the layer is a 16-kernel 32-by-
32-by-16 matrix. The output of the neurons in the first is
calculated using Equation (4):
Sa = ag M lg Pactierseteesesinase (4)
where Sy characterizes the output feature maps, Cy
epitomizes the maps, and T denotes the weighted map.
After that, the layer's output is transformed into an energy
descriptor. After the descriptor. It performs similarly to a
texture explanation for a cluttered, thick surface. In
Equation (5), we see the relationship stated as:
BILE AO Ol ay Ee OEP | wetexveeaspereesee: (5)
where EL(é,0) characterizes the EL layer, j
characterizes the input, and T signifies the EL weighted
vector. The link between the EL and FC layers is
substantially smaller compared to the parameters. In
addition, EL remembers the energy state of the previous
layer and acquires new knowledge as signals travel both
forward and backwards in time. In addition to enhancing
the learning capacity and simplifying the projected
system, EL helps lower the vector scope of the following
FC layer. To determine which EL parameters may be
taught, use Equation (6):
Se a scaeaiweeeeares (6)
where &,, is the EL learnable limits, 7” is FC layer
neuron, and 7~+ is the preceding FC layer neuron.
Between the convolution layer and the rectified linear
unit (ReLU) layer, a batch normalization and activation
function is utilized to expedite the shift, which can be
eliminated by normalization. The deviation can be
normalized to accomplish this. Mean and Variance are
strongminded using Equations (7) and (8) used in the
bulk normalization computation.
1
Tq = oe Li
OT (eee) eee ere (8)
DOI: https://doi.org/10.52756/ijerr.2023.v33spl.003
where Tg and vg characterize the mean and alteration
correspondingly, n is the size of 1; element of features.
Normalization is intended in Equation (9) as:
Oj-T
A, = jaa Te Aniolaiaabiaclde ateutied (9)
Where a and A are the two starting parameters of the
output layers that can be learned. The activation function
for the ReLU layer can be found in Equation (10) and its
output can be found in Equation (11), as:
AijK = max0, Dijk i faeleetaie tte | eee (10)
Arety = ReLU(Bnorm(Conv(w,x))) ......1)
Where A; ;, represents the output features and ¥; ; ,
symbolizes the feature of the input element. Afterwards,
The control network is over fit because the pooling layer
averages out data from the feature maps, weights, and
computations. The formula is as follows, and it is used to
determine the maximum pooling layer:
Miao = ae (0,20 Tp) caste 12)
Where Myo; signifies the production feature maps, 3
designates the maps, Q means the pooling size, and T
stands for the maximum pooling layer for the kernel
vector. In this study, we employ a maximum of two
pooling layers, each having a kernel size of two by two.
To avert the model from overfitting the training data,
the layer is utilized during the weighted update phase to
repeatedly eliminate a sample of accidental parameters.
To prevent overfitting training data, drop editing is
performed during the weighted update phase to
periodically eliminate a subset of random parameters.
Over-compatibility of training data is _ especially
problematic in FC layers since they include the most
network-wide properties. The dropout layer is the result
since the FC layer is established later. The softmax layer
is used as a loss-based classifier. Softmax accepts and
one, between [0, 1]. In Equation (13), the loss function is
expressed mathematically as:
Kp = Oj + 10g Di OXP (02) onivicavas vores (13)
Where k; signifies the entire loss and 6; consuming
the class d which is i-th course element. The classifier's
goal is to reduce the likelihood between the true label and
its projected counterpart, as computed by the function in
Equation (14):
Int. J. Exp. Res. Rev., Vol. 33: 18-29 (2023)
_ expi
i Yiexp(6j)
The next step of HPT-TCNN, hyper-parameter
tweaking, is described below once this phase is complete.
Input and output dimensions for the proposed network are
listed in Table 1.
Table 1. Projected HPT-TCNN structure layers
In Eq. (16), ByR, Fol, Ovr, and Att are the “bypass
rider, supporter, overtaker and attacker’’,
correspondingly. Furthermore, parameters which are
cluster loading. The steering angle of the vehicle at a
period is given in Eq. (17), which S jae is the angle of c™
Kernel Size to Form
Input Size Padding Respectively Feature Stride Output Size
EL 16x16x64 - - - 128x1
Dropout 128x1 - - - 128x1
FC1 128x1 - - - 1024x1
Dropout 1024x1 - - - 1024x1
FC2 1024x1 - - - 2x1
Convolutional Layer 1 64x64x1 [1111] 5x5 [1 1] 62x62x16
Max Pooling Layer 1 62x62x16 [1111] 2x2 [2 2] 32x32x16
Convolutional Layer 2 32x32x16 [1111] 5x5 [1 1] 30x30x32
ReLU
Max Pooling Layer 2 30x30x32 [1111] 2x2 [2 2] 16x16x32
Convolutional Layer 3 16x16x32 [1111] 3x3 [1 1] 16x16x64
ReLU
Classification Layer - - -
Hyper-parameter tuning process using MGS-ROA
The cluster of riders is progressing toward the goal,
which inspired the algorithm ROA (Binu and Kariyappa,
2019). Let's pretend many groups of cyclists are headed
in the same general direction. The cyclists are divided
into four groups, each including an equal number of
riders. Bypass riders follows, overtakers and attackers
make up the four groups of riders. Each faction has its
unique strategy to reach its goal. The rider attempting to
pass the leader does so by focusing on the leader's
position relative to his own and then moving in that
direction. The assailant arrives quickly and stands in the
rider's path to the objective. Each rider should follow the
steps outlined below, which comprise this algorithm.
Group and rider parameters are set to their default
values. Riding groups (RG) are first set up with a random
distribution of riders among four groups. Eq. (15) is a
representation of the cluster initialization. Here, the
number of riders (RN) is synonymous with the riders'
group (RG). The coordinate number (CN) represents the
number of dimensions. Lit (c, d) also represents the
location of the c™ rider. The total number of riders may
be determined by adding up the riders in each section,
and the related equation is given by the symbol Eq. (16).
Lit = (Liz, 1S cS RN; 1 Sd
% 0.920 4
5
u
= 0.915 4
0.910 4
0.905 4
0.900 + T T T T T
0 5 10 15 20 25
Epochs
Figure 3. Accuracy graph investigation of Proposed procedure
Training and Validation Loss
— Training
023-4 —— Validation
0.22 4
0.214
a
S 0.204
0.19 4
0.18 5
15 20 25
Epochs
Figure 4. Loss graph investigation of the proposed procedure
Performance metrics
We utilized the measures to gauge how well our
model performed. Equation (33) may be used to
determine the precision:
TP+TN
TP+TN+FP+FN
The four outcomes are a True Positive (TP), True
Negative (TN), a False Positive (FP), and a False
Negative (FN).
The precision is intended by the subsequent Equation
(34):
Accuracy =
Precision = ——
DOI: https://doi.org/10.52756/ijerr.2023.v33spl.003
TP
TP+FN
The F1 is designed by the subsequent Equation (36):
Recall =
F1l= ae
Precision+recall
The AUC curves evaluate the false positive and true
positive rates at various cut-offs. Precision at each
threshold is weighted equally in AP's summary of the
curve's recall. Accuracy and loss of the projected model
on training and testing data are shown in Figures 3 and 4,
respectively.
Int. J. Exp. Res. Rev., Vol. 33: 18-29 (2023)
Table 2. Analysis of the proposed model for binary classification
Class Labels Accuracy Precision Recall Specificity
Training Set (70%)
Benign 92.10 92.64 91.39 92.80 92.01
Malignant 92.10 91.58 92.80 91.39 92.18
Average 92.10 92.11 92.10 92.10 92.10
Testing Set (30%)
Benign 93.60 93.95 93.33 93.88 93.64
Malignant 93.60 93.25 93.88 93.33 93.56
Average 93.60 93.60 93.60 93.60 93.60
Benign
Malignant
Disease Types
m Accuracy Precision Recall sSpecificity m= F-Score
Figure 5. Graphical analysis of a proposed model for training set
In the Training Set (70%) of splitting condition, the
Benign reached an accuracy of 92.10, a precision value of
92.64, a recall value of 91.39, and a specificity of 92.80
and the Fl-score of 92.01, respectively. Malignant
reached an accuracy of 92.10, a precision value of 91.58,
a recall value of 92.80, a specificity of 91.39 and an F1-
score of 92.18, respectively. The average value reached
an accuracy of 92.10, a precision value of 92.1, a recall
value of 92.10, and a specificity of 92.10 and the FI-
score of 92.10, respectively. After the Testing Set (30%)
DOI: https://doi.org/10.52756/ijerr.2023.v33spl.003
condition, the Benign reached an accuracy of 93.60, a
precision value of 93.95, a recall value of 93.33 and a
specificity of 93.88 and the Fl-score Of 93.64,
respectively. Malignant reached an accuracy of 93.60, a
precision value of 93.25, a recall value of 93.88, a
specificity of 93.33, and an Fl-score of 93.56,
respectively. The average reached an accuracy of 93.60,
precision value of 93.60, recall value of 93.60, and
specificity of 93.60, respectively.
Int. J. Exp. Res. Rev., Vol. 33: 18-29 (2023)
Benign Malignant
Disease type
mAccuracy Precision Recall m Specificity mF-Score
Figure 6. Analysis of the proposed model for testing set
= Accuracy ©&Sensitivity & Specificity
MLP CNN Proposed
Models
Figure 7. Graphical comparison for various DL models
The comparative analysis of different DL Models is
shown in Table 3. The AE model analysis's accuracy,
sensitivity, and specificity were 0.8311, 0.8923, and
AE 0.8301 0.8923 0.8242 0.8242, respectively. After that, the DBN model attained
Table 3. Comparative analysis of various DL models
Methods Accuracy Sensitivity Specificity
accuracy and sensitivity values of 0.8416 and 0.8631,
DBN 0.8416 0.8631 0.8715
respectively. The ELM model achieved an accuracy of
ELM 0.8722 0.8428 0.9016 0.8722, a sensitivity of 0.8428, and a specificity of
MLP 0.8880 0.9020 0.8676 0.9016. The MLP model then achieved an accuracy of
CNN 0.8979 0.9293 0.8358 0.8880, sensitivity of 0.9020, and specificity of 0.8676,
respectively. The accuracy, sensitivity, and specificity
Proposed 0.9360 0.9360 0.9360 values for the CNN model were then 0.8979, 0.9293, and
0.8358, respectively. Finally, the suggested model
DOE: https://doi.org/10.52756/ijerr.2023.v33spl.003
Int. J. Exp. Res. Rev., Vol. 33: 18-29 (2023)
achieved the following results: accuracy of 0.9360,
sensitivity of 0.9360, and specificity of 0.9360.
Conclusion
In order to increase the precision and effectiveness of
Monkeypox detection, this study presents a unique
strategy that combines several approaches. The process
begins by enhancing the original Monkeypox photos by
Next,
extracted using an energy layer. Then, the Hyper-
parameter-based transferable convolutional neural
network (HPT-TCNN) is introduced to improve
classification performance even further. Notably, this
enhancing edge detail. texture features are
method streamlines the process by substituting just three
convolutional layers and one energy layer for the
conventional pooling layer. The model is more
approachable and effective because of the usage of the
Optimisation Algorithm (MGS-ROA), which makes
hyper-parameter adjustment easier.
Additionally, the "Monkeypox Skin Lesion Dataset
(MSLD)" was developed by gathering pictures of
Monkeypox-related skin lesions on people. This database
is a valuable tool for training and validation. The
comparison of several deep learning models highlights
the suggested model's higher performance. It performs
better than models like AE, DBN, ELM, MLP, and basic
CNN with accuracy, sensitivity, and specificity, all at
93.60 percent. The model's exceptional accuracy, well-
balanced sensitivity, and specificity values indicate how
well it can categorize Monkeypox skin lesions. This
research reveals a novel method for detecting
Monkeypox and offers convincing proof of its higher
efficacy compared to other models. For doctors and other
healthcare workers, its application offers enormous
potential since it can speed up early detection and
eventually help limit Monkeypox outbreaks.
Funding Statement
The authors declare that no funds, grants, or other
support were received during the preparation of this
manuscript.
Competing Interests
The authors have no relevant financial or non-
financial interests to disclose.
Data Availability
Data sharing does not apply to this article as no
datasets were generated or analyzed during the current
study.
DOI: https://doi.org/10.52756/ijerr.2023.v33spl.003
References
Adalja, A., & Inglesby, T. (2022). A Novel International
Monkeypox Outbreak. Annals of Internal
Medicine, 175(8), 1175-1176.
https://doi.org/10.7326/M22-1581.
Altun, M., Giiriiler, H., Ozkaraca, O., Khan, F., Khan, J.,
& Lee, Y. (2023). Monkeypox Detection Using
CNN with Transfer Learning. Sensors (Basel),
23(4), 1783. https://doi.org/10.3390/s23041783.
Almufareh MF., Tehsin S., Humayun M., & Kausar S.
(2023). A Transfer Learning Approach for Clinical
Detection Support of Monkeypox Skin Lesions.
Diagnostics (Basel), 13(8), 1503.
https://doi.org/10.3390/diagnostics 13081503.
Ariansyah, M. H., Winarno, S., & Sani, R. R. (2023).
Monkeypox and Measles Detection using CNN
with VGG-16 Transfer Learning. Journal of
Computing Research and Innovation, 8(1), 32-44.
https://doi.org/10.24191/jcrinn.v8i1.340
Ayca, A. K., Vedat, T., Ipek, M. (2022). Motor imagery
EEG signal classification using image processing
technique over GoogLeNet deep learning
algorithm for controlling the robot manipulator.
Biomedical Signal Processing and Control, 72(A),
103295, ISSN: 1746-8094.
https://doi.org/10.1016/j.bspc.2021.103295.
Binu, D., & Kariyappa, B. S. (2019). RideNN: A New
Rider Optimization Algorithm-Based Neural
Network for Fault Diagnosis in Analog Circuits.
IEEE Transactions on _ Instrumentation
Measurement, 68(1), 2-26.
https://doi.oeg/10.1109/TIM.2018.2836058.
Catala, A., Clavo-Escribano, P., Riera-Monroig, J.,
Martin-Ezquerra, G., Fernandez-Gonzalez, P., ...
& Fuertes, I. (2022). Monkeypox outbreak in
Spain: clinical and epidemiological findings in a
and
prospective cross-sectional study of 185 cases. The
British Journal of Dermatology, 187(5), 765-772.
https://doi.org/10.1111/bjd.21790.
Chollet, F. (2018). Keras: The python deep learning
library. Astrophys. Source Code Library, Tech.
Rep. Rec. Ascl:1806.022.
Jadhav, A. S., Patil, P. B., & Biradar, S. (2021). Optimal
feature selection-based diabetic retinopathy
detection using improved rider optimization
algorithm enabled with deep learning.
Evolutionary intelligence, 14(4), 1431-1448.
https://doi.org/10.1007/s 12065-020-00400-0
Khalifa, N. E., Loey, M., & Mirjalili, S. (2022). A
comprehensive survey of recent trends in deep
learning for digital images augmentation. Artificial
Int. J. Exp. Res. Rev., Vol. 33: 18-29 (2023)
Intelligence Review, 55(3), 2351-2377.
https://doi.org/10.1007/s10462-021-10066-4.
Li, D. & Ling, D. (2021). Recent advances of deep
learning algorithms
vision systems with emphasis on fish. Artificial
55(4).
for aquacultural machine
Intelligence Review.
https://doi.org/10.1007/s10462-021-10102-3.
Nolen, L. D., Osadebe, L., Katomba, J., Likofata, J.,
Mukadi, D., & Reynolds, M. G. (2016).
Extended Human-to-Human Transmission during a
Monkeypox Outbreak in the Democratic Republic
of the Congo. Emerging Infectious Diseases, 22(6),
1014-1021.
https://doi.org/10.3201/e1d2206. 150579.
Out, A., Ebenso, B., Walley, J., Barcelo, J. M., & Ochu,
C. L. (2022). Global human Monkeypox outbreak:
atypical presentation demanding urgent public
health action. The Lancet, Microbe, 3(8), e554-
e555. https://doi.org/10.1016/S2666-
5247(22)00153-7.
Pramanik, R., Banerjee, B., Efimenko, G., Kaplun, D., &
Sarkar, R. (2023). Monkeypox detection from skin
lesion images using an amalgamation of CNN
models aided with Beta function-based
normalization scheme. PLoS One, 184),
e0281815.
https://doi.org/10.1371/journal.pone.0281815.
Samaranayake, L., & Anil, S. (2022). The Monkeypox
Outbreak and Implications for Dental Practice.
International Dental Journal, 72(5), 589-596.
https://doi.org/10.1016/j.identy.2022.07.006.
Sanaat, A., Shiri, I., Ferdowsi, S., Arabi, H., & Zaidi H.
(2022). Robust-Deep: A Method for Increasing
Brain Imaging Datasets to Improve Deep Learning
Models’ Performance and Robustness. Journal of
Digital Imaging, 35, 469-481.
https://doi.org/10.1007/s10278-021-00536-0.
How to cite this Article:
Serkan. (2022).
Disease
Detecting the Stages of
with Pre-trained Deep
Arabian Journal for
Science and Engineering, 47(2), 2201-2218.
https://doi.org/10.1007/s13369-021-06131-3.
Suganyadevi, S., Seethalakshmi, V., & Balasamy, K.
(2022). A review on deep learning in medical
Savas,
Alzheimer’s
Learning Architectures.
image analysis. International Journal of
Multimedia Information Retrieval, 11(1), 19-38.
https://doi.org/10.1007/s13735-021-00218-1.
Uysal, F. (2023). Detection of Monkeypox Disease from
Human Skin Images with a Hybrid Deep Learning
Model. Diagnostics (Basel), 13(10), 1772.
https://doi.org/10.3390/diagnostics 13101772.
Yasmin, F., Hassan, M. M., Hasan, M., Zaman, S.,
Kaushal, C., El-Shafai, W., & Soliman, N. F.
(2023). PoxNet22: A fine-tuned model for the
classification of Monkeypox disease using transfer
learning. IEEE Access, 11, 24053-24076.
https://doi.org/10.1109/ACCESS .2023.3253868.
Yue, X., Li, H., Shimizu, M., Kawamura, S., & Meng, L.
(2022). YOLO-GD: A Deep Learning-Based
Object Detection Algorithm for Empty-Dish
Recycling Robots. Machines, 10(5), 294.
http://dx.doi.org/10.3390/machines 10050294
Zhang, J., Li, C., Rahaman, M. M., Yao, Y., & Ma, P.
(2021). A comprehensive review of image analysis
for microorganism counting: from
classical image processing
approaches. Artificial Intelligence Review, 55(4),
2875-2944. https://doi.org10.1007/s10462-021-
10082-4.
Zumla A., Valdoleiros, S.R., Haider, N., Asogun, D., &
Ntoumi, Petersen, E., & Kock, R. (2022).
Monkeypox outbreaks outside endemic regions:
priorities. The Lancet
Infectious Diseases, 22(7), 929-931.
https://doi.org/10.1016/S 1473-3099(22)00354-1.
methods
to deep learning
scientific and social
V. Gokula Krishnan, B. S. Liya, S. Venkata Lakshmi, K. Sathyamoorthy and Sangeetha Ganesan (2023). Monkeypox detection using
hyper-parameter tuned based transferable CNN model. International Journal of Experimental Research and Review, 33, 18-29.
DOI: https://doi.org/10.52756/ ijerr.2023.v33sp1.003
«c) MAE) This work is licensed under a Creative Commons Attribu-
ance tion-NonCommercial-NoDerivatives 4.0 International License.
DOI: https://doi.org/10.52756/ijerr.2023.v33spl.003