SKIN CANCER DETECTION AND SEVERITY PREDICTION USING DEEP LEARNING

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 11 | Nov 2023 www.irjet.net p-ISSN: 2395-0072
© 2023, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 543
SKIN CANCER DETECTION AND SEVERITY PREDICTION USING
DEEP LEARNING
Sasikala G M1, Joshikha V2, Keerthana E3
1 Assistant Professor, Department of Computer Science and Engineering, Rajalakshmi Engineering College,
Tamil Nadu, India
2 Undergraduate Student, Department of Computer Science and Engineering, Rajalakshmi Engineering College,
Tamil Nadu, India
3 Undergraduate Student, Department of Computer Science and Engineering, Rajalakshmi Engineering College,
Tamil Nadu, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Melanoma isamalignancythatisfatal. Melanoma
scenarios are more frequent in Northern India, whereas non-
melanoma skin cancer cases are more common in the
Northeast, especially in Nagaland, according to a new
research by the Indian Council of Medical Studies (ICMR). Skin
melanoma is anticipated to be the cause of over 7650
accidental deaths in 2022, according to a US Cancer Institute
analysis. All around the world, the age group of 70 years and
above age interval is most impacted by melanoma. The
suggested approach may identify cancer types that are
melanocyte, squamous cell, and basal cell. We give two
approaches to a unique problem: cross-domain skin disease
identification. Using two datasets pertaining to skin
conditions, the network was refined.
Key Words: Deep Learning, Melanoma, ImageNet,
Transfer Learning, Convolutional Neural Networks.
1.INTRODUCTION
Approximately 20 square feet make up the skin, which isthe
greatest component of the human body. The primary
functions of skin include assisting the body in controlling
body temperature, shielding interior organs from UV
radiation and bacteria, and enabling touch, heat, and cold
sensations. Recent years have seen a sharp rise in the
prevalence of skin cancer. Most fatalities from skin cancer
are caused by melanoma; the World HealthOrganization,for
example, predicts that232,000casesofmelanoma havebeen
documentedworldwide.Furthermore,theagedifference and
rate are said to be rising globally each year. With higher
possibilities of long-term survival, early discovery makes
treatment crucial. The survival ratedropsdramaticallyandit
gets more difficult to threaten if it is not discovered early.
The melanocyte, Squamous, and basal are the three primary
groups into which skin cancer can be divided. Typically,
basal cell carcinoma, the most common kind of cancer,
develops very slowly and does not metastasize to other
regions of the body. HPV's imperative to eradicateHPVfrom
the body since it resurfaces often. Squamous cell carcinoma,
also called squamous cell is another form of skin cancer that
is more likely to progress to other areas of the body and
penetrates the skin far more deeply than basal cell
carcinoma. Melanocytes, the last class of cells, produce
melanin in reaction to sunlight, which is responsible for the
tan or brown color of skin. The melanin in these cells
protects the outermost layer of skin from sunlight, but an
excess of it can accumulate inside and result in malignant
moles or cancer calledmelanoma.While basal andsquamous
cancers are generally considered to be benign because they
tend to cause relatively little harm to the surrounding
tissues, melanocyte-based malignant tumors have been
deemed malignant and can be lethal.
When melanoma is discovered or diagnosed in its initial
stages, it can be cured; however, if it is discovered or
diagnosed in its later stages, it may spread deeper into the
layer of skin and affect various regions of the body, in which
case treatment may become extremely challenging. Because
melanocytes are found throughout the body, melanoma is
brought forth by them. A key contributing factor to the
development of melanoma is skin exposure to UV light.
Melanin deposits are often located in the layer of epidermis
in benign lesions. The amount of melanin generated in
malignant tumors is very aberrant. Until melanocytes and
the melanin they produce stay in the epidermal covering,
malignant lesions are not life-threatening; nevertheless, the
color of the skin changes as soon as the melanocytes reach
the dermal layer along with leave sediments behind. The
British Skin Foundation estimates that each year, over
100,000 new instances of cancer of theskinaredetected and
that the illness claims the lives of about 2500 individuals.
Lots of individuals worldwide struggle with skin cancer,
which is becoming more common. In general, this poses
issues to the global health community. Up until theageof 70,
one in five Americans will have skin cancer. Each year in
Europe, melanoma causes the deaths of 22,000 individuals
and confirms over 100,000 new cases of the disease.
However, one of the most amazing thingsaboutthecancer of
the skin is that the five-year recovery probability increases
to 99% when the cancer is diagnosed early, compared to a
23% likelihood when it is discovered later in the disease's
course.

Fig-1: Types of skin cancer
As a result, it is crucial to diagnoseskincancerearly.
Dermatologists can identify skincancerbya straightforward
visual inspection of lesions. Even for experienced medical
professionals, distinguishing between benignandmalignant
skin conditions can be challenging because of the subtle
differences between them. As a result, it is good to have
medical applications in this sector that offer robotic skin
lesion assessment for decision advice. As prospective tools
for the assessment of melanocytic lesions, computer-aided
evaluation (CAD) systems are being explored extensively
since the early 1990s. The computerized examination was
initially carried out using preset methods that were well-
known to dermatological experts, such as the ABCD-rule.
However, these methods frequently failed both to
extrapolate to new incidents or to achieve the necessary
accuracy. In order to provide seasoned physicians a second
opinion or to provide novice physicians with enhanced
training, a number of software programs for the automatic
diagnosis of melanoma insynoptic anddermoscopic pictures
are now available on the market.
To improve both the rapidity and precision of
diagnosis, the automatic technique can be highly beneficial.
This makes it increasingly important to build algorithms for
illness diagnostics that identify and use different elements
from the photos. The diagnosis of illnesses is one of the
numerous uses for computational imaging that has recently
attracted a lot of empirical attention.
2.LITERATURE SURVEY
Research on the diagnosis of skin cancer, its severity, and
different methods for classifying skin cancer disorders is
offered by a number of writers and is summarized here.
[1] The primary goal of this study is to develop a system
that can automatically determine if a skin cancer case is
benign, a malignant tumor or neoplastic. In this study, they
made use of the following methods: feature extraction,
segmentation, noise reduction, picture capture, and
preliminary processing. For the very initial a period of time
Cubic Regression has been used in GuidedMachineLearning
techniques. We were able to teach the machine to identify
the different stages of skin cancer and to show Benign,
Melanoma, and Melanoma automatically by using this
approach. Overfitting: Given that the training setistoolarge,
the model performs well on it but cannot generalize to
unfamiliar, unproved information.
[2] The goal of this study is to determine the best accurate
skin cancer diagnosis and to classify melanoma as non-
malignant or cancerous. A few pre-processing steps,
including as the process of segmentation glare, a shadow,
and removal of hair, were finished in order to accomplish
this. Deep Neural Networks and Support Vector Machines
will be utilized for classification. Before being utilized for
classification, the classifier will be taught to understand the
characteristics. The present approach is novel inthatitsgoal
is to detect abnormalities rapidly, which will help
technicians improve theirdiagnosticskills. TheInternational
Standards Institute (International SkinImagingConsortium)
dataset is openly accessible, therefore any collection of
images may be used to calculate the efficiency.
[3] The most effective techniques are used in this study to
categorize all forms of cancer. The dull razor technique, the
Gaussian filter, and the median filter are all employed to
improve pictures and eliminatenoise.Featuresare extracted
from the segmented pictures using two methods: the GLCM
methodology and the ABCD method of feature extraction
procedure. Features from both strategies are combined to
enhance categorization. In the end, a The Microsoft Virtual
Machine classifier is used for classification inordertoobtain
high accuracy.
[4] This work examines how Deep Learning Studio (DLS)
may be used to apply Model Driven Approach to the field of
deep learning. It demonstrates DLS's capabilitiesanddetails
how to generate a Deep Learning System using the tool. This
work demonstrates the process of preparing dermal cell
pictures for information entry and identifies cancer cells
using the DLS model. The outcomes demonstrate good
accuracy, with DLS models achieving an Area Underneath
the Curve (AUC) of 99.77% in recognizing malignant cells
from the pictures.
[5] Train and assess the system, the study made use of three
databases: PAD-UFES-20, an image subset from
Fitzpatrick17k, and the ISIC Dermoscopic Archive for the
2019 and 2020 melanoma detection challenges. For lesion
classification, Convolutional artificial neural networksofthe
Inception-Re kind were employed. The suggested approach
demonstrated a 94.5% accuracy rate in differentiating
between benign and malignant lesions and an 89.3%
accuracy rate in classifying four common skin conditions.
[6] The 11,527-image ISIC 2018 dataset was used for
training and testing of the proposed technique. Prior to
applying a CNN model to the test set, processes for
preprocessing such data augmenting, scaling, normalizing,
and picture enhancement employing ESRGAN were
deployed. The system applied image advancement

algorithms on lesion pictures to boost the visibility and
minimize noise. In order to minimize overfitting and
enhance the overall effectiveness of the suggested deep
learning approaches, a number of models, namelyResnet50,
InceptionV3, along with Resnet Inception's design were
trained on preprocessed lesion medical imagery.
[7] Dermoscopic skin lesion images are used by the
proposed model to classify skin cancer cases. Resnet-
50/CNN is used for preprocessing, classification, and
training in the operational approach. To enhance image
quality, the first preprocessing step is carried out using
ESRGAN. Ground truth pictures are used to establish the
area of interest (also known as ROI) for each segmented
tumour. Then, for real-time development and skin cancer
grouping, dermoscopy photos are supplied into the Resnet-
50/CNN networks.
[8] In order to detect melanoma more effectively, the study
presents a novel deep network architecture called Xception.
Using depthwise separable convolutions and the Swish
activation function, the model improves CNN's classification
accuracy. Having a 95.53%F1-scoring94.05%sensitiveness,
97.07% exactness, and an accuracy of 100%, the suggested
method outperformed most advanced approaches when it
came to the MNIST skin cancer dataset.
[9] An approach for segmenting non-uniform images using
level-set segmentation is presented in the study. The
recommended approach has a mistake percentageof3.39%,
an incidence of false positives at 3.62%, and an accurate
finding rate of 94.4%. The method's ability to precisely
segment lesions while being unaffected by variations in
brightness, contrasts, hairs, and blood vessels is confirmed
by the testing results.
[10] By adding uncertainty, fuzzy clustering lowers the
number of false positives in models. Variations in image
quality, such as lighting, resolution, and artifacts in
dermoscopic images, can have an impact on the accuracy of
the model.
[11] This paper provides a summaryoftheadvancesmade in
deep learning techniques for the categorisation of skin
cancer. Images from various dermatological disciplines as
well as frequently utilized datasets are provided.Theuses of
CNN-based approaches for the classification of skin cancer
will be addressed in the article, including frontier problems
like data disparities, cross-domain adapting, stability, and
profitability. It additionally supplies relevant methods that
utilize deep learning to tackle these sorts of problems.
[12] The created neural network architecture can help
dermatologists with melanoma screening, according to the
study's multiclass classification results. Comparisons with
other approaches show how reliable and effective the
method is. Real-time skin melanoma detection can be
effectively addressed by the suggested approach, which can
be integrated into low-resource devices and yieldpromising
results.
[13] In the paper, a deep neural network-based smartphone
app for diagnosing melanoma is presented. An average of
89% balanced accuracy and 100% recall are obtained from
clinical data. Furthermore, the outcomes produced with a
weighted loss function and a mutation operator inspired by
DE closely match each other.
[14] AI might be a big help in the detection of skin cancer.
The two primary areas of artificial intelligence (AI) utilized
in the identification and categorization of cancers oftheskin
are shallow and deep approaches. This study shows that
accuracy results are greater when fewer testing classes are
included. Our data also shows that, in contrast to common
assumption, greater accuracy ratings have been observed
when fewer observations are included. This might be due to
a number of things, such as the kind of photographs and the
methods used. Further, every AI tool must first undergo
autonomous outside validation on a large, varied, and
unbiased database in order show its reliability and
universality before it is used in a clinical environment.
[15] In this work, a Rapid Region-based neural network
using convolution (FRCNN) is used to comparetheprecision
of classification of pigmented skin lesions withthediagnosis
accuracy of dermatologists. Using 5846 clinical photographs
from 3551 recipients, the FRCNN achieved unpaid results:
86.2% accuracy in six-class category and 91.5% confidence
in the malignant versus benign classification. A comparison
review revealed that FRCNN performed better in terms of
accuracy and specificity than 10 board-certified
dermatologists and trainees. Themethodhasthepotential to
be widely adopted by society, which might significantly
enhance skin cancer prognoses and highlight the
revolutionary potential of computer science in medical
evaluations.
[16] The difficulty of diagnosing skin conditions—especially
skin cancer—is discussed in this literature review, which
also highlights the promise of big data healthcare
frameworks for accurate prediction and early diagnosis in
the future. The effectiveness of the fuzzy Analytical
Hierarchy Process (AHP) technique in summarizing and
forecasting is emphasized. The essay suggests a big data
technology-based approach toclassifyskincancertreatment
phases. Fuzzy selection experiments using a support vector
machine for classification with multiple classes and binary
migration classification based on the radial basis function
demonstrate the effectiveness of the approach, exceeding
rivals with 90.86% effectiveness. By assisting in the
classification of skin cancer severity and the identification of
mass phases, the methodology illustrates the approach's
possible clinical relevance.
[17] This literature review highlights the rising global
incidence of skin tumors and the critical need for prompt

detection. A number of types, including deadly ones like
melanoma, are caused by genetic and metabolic
abnormalities. A specific type of deep learning called
Convolutional neural networks, more commonly has shown
potential as a skin cancer detection tool. For precise skin
lesion classification, the study suggests a CNN architecture
consisting of five layers that uses diagnosis labels and image
pixels as inputs. To assess tumor intensity after
classification, a different approach makes advantage of
OpenCV color prediction. This approach has the potential to
enhance early detection and ultimately save lives by
leveraging the capabilities of state-of-the-art deep learning
algorithms in medical picture analysis.
[18] This review of the literature explores the difficulties in
diagnosing melanoma, highlighting the disease's rising
global incidence and the vital importance of early
identification. With the use of deep learning, image
processing, and machine intelligence—more especially,
convolutional neural networks with deep layers, or CNNs—
the study explores the automated detection of carcinoma
from dermoscopic skin samples. In comparison to other
models, the suggested DCNN model—a hybridofVGG16and
LSTM—performs better. The model attains remarkable
testing and training accuracies of 94.39% and 90.89%,
respectively, through rigorous preprocessing and
regularization techniques. This work demonstrates how
state-of-the-art deep learning techniquesmaytransform the
diagnosis of melanoma and enhance the lives of patients.
[19] The worldwide impact of skin cancer is emphasized by
this literature review, whichalsohighlightshowcrucial early
detection is to effective control. The shortcomings of the
severity detection algorithms used today include large time
and cost commitments as well as less-than-ideal lesion
analysis. (MPNN) classifier is proposed to overcome these
difficulties by using Spider Monkey Optimization for neural
network training. The classifier performs better than more
well-known methods like KNN, among others NBandOBRF,
and SVM having a 0.10% false- positive Rates, 0.03%
mistakes, and 0.98% confidence. This study demonstrates
how the MPNN classifier can improve skin cancer severity
detection, offering a more precise and effective diagnostic
method.
[20] This literature review suggests a Graph Convolutional
Network (GCN) with Delaunay triangulation for feature
extraction to address the crucial problem of skin cancer
misdiagnosis. Boundary extraction is aided by Delaunay
triangulation, which enables the model to concentrate only
on malignant lesions for more precise predictions. With
benefits over conventional CNN models, GCN is particularly
good at simulating interactions between image regions and
structures and enables node-to-node messaging. Using few-
shot and transfer learning, GCN handles a small number of
annotated medical image datasets. But problems like
overfitting occur, whicharecausedbythingslikeunbalanced
data, inaccurate feature extraction, and not enoughfeatures.
The study offers insightful informationaboutimprovingGCN
models for more accurate skin cancer classification.
3.PROPOSED METHOD
Using deep learning and image processing, a skin disease
detection and classificationsystemcanbedevelopedinthree
main steps. Create a diverse dataset of annotated images of
skin diseases first. Preprocess the photographs by resizing,
establishing a normal and improving them to increase the
dataset's variability. Lastly, use a deep learning building
design, such a neural network based on convolution (CNN),
to train the model utilizing the preprocessed data. As two
rigorous evaluation criteria, use exactness and F1-score.
Backpropagation should be used to adjust these parameters
and optimize the model. Apply extraction of attributes and
methods of augmentationto imagestoguaranteerobustness.
4.IMPLEMENTATION
The recommended approach is shown, and every section is
explained in greater detail below. input The proposed
approach utilizes databases of high-resolution dermoscopic
images. Using 800 compressed pictures from the
International Standards Institute 2019 challenge dataset—
which consists of eight different classes—the proposed
system.
Pre-processing: Several factorsneedthepictureacquisition
process to be non-uniform. Consequently, the main goal of
the pre-processing phase is to reduce or remove the
backdrop or other unwanted portions of the image in order
to improve the image properties, such as quality, clarity, etc.
The three main preliminary processing stages are grayscale
conversion, noise reduction, and picture enhancement. All
the images in the proposed system are first transformed to
grayscale. The Gaussian mask and a median filter are then
used for noise reduction and picture enhancement. To
remove undesirable hair from the area of lesion, filters are
employed in addition to the Dull Razor Method. The

improvement of images aims to increasean image'svisibility
while simultaneously enhancing its quality. The bulk of skin
lesions tend to consist of skin cells and hair, which could
render achieving excellent classification precision difficult.
The formula for Gaussian filter is,
Consequently, the excess hair from the photographs is
removed using the dull razor procedure.
Thus, reconstruction loss is defined as follows:
The Dull Razor techniqueis mostlyusedtoaccomplish
these tasks: A) Using this grayscale morphological
operation, the location of the hair on the outer layer of skin
lesion is determined. b) The hair pixel is updated using
bilinear interpolation once the position has been
determined and the shape whether it be long or thin has
been verified.
Fig-2: Image from dataset (left); enhanced image
(right)
Segmentation: Divide the region that is relevant in a
picture by using the segmentation method. One way to
achieve this separation is to assign a similar attribute to
every pixel in the image. This has the primary benefit that
separated images can be processed separately from the
entire image. Identifying the boundaries of the specific
area is the most widely used method. Detecting
similarities within a specific region is the basis for other
techniques like thresholding, clustering, and region
growing.
Fig-3:Segmented image
Feature extraction: In the entire classification process,
feature extraction is thought to be the most important step.
Feature extraction is the process of removing pertinent
features from a provided input dataset so that they can be
used in subsequent computations like further detection and
classification.
5.CONCLUSION
Diagnosing skin cancer is a challenging assignment for
dermatologists because many skin cancer colors aresimilar.
Therefore, early detection of lesions is essential and helpful
for providing patients with complete recovery from skin
cancer. Deep convolutional neural networks(DCNN)models
have been widely studied for the aim of identifying skin
disorders. In several cases, these simulations have even
yielded diagnostic results that arecomparabletoorsuperior
than those obtained by dermatologists. The use of
appropriate computational layers, network depth,
hyperparameter tuning, and the selection of various filters
and size choices are crucial aspects of DCNN design. The
limited quantity and unbalanced data of publicly accessible
skin lesion datasets, however, limit the use of DCNN in the
detection of skin disorders. This study presents anoverview
and comparison of many types of CNN.
REFERENCES
[1] Gaana, M and Gupta, Shweta and Ramaiah, Narayana
Swamy, Diagnosis of Skin Cancer Melanoma using Machine
Learning (March 22, 2019).
[2] M, Vijayalakshmi. (2019). Melanoma Skin Cancer
Detection using Image Processing and Machine Learning.
International Journal of Trend in Scientific Research and
Development. Volume-3. 780-784. 10.31142/ijtsrd23936.
[3] Monika, M. & Vignesh, N. & Kumari, Usha & Kumar,
M.N.V.S.S. & Lydia, Laxmi. (2020). Skin cancer detection and
classification using machine learning. Materials Today:
Proceedings. 33. 10.1016/j.matpr.2020.07.366.
[4] Mohammad Ali Kadampur , Sulaiman Al Riyaee,Skin
cancer detection: Applying a deep learning based model
driven architecture in the cloud for classifying dermal cell
images ,Informatics in Medicine Unlocked,Volume18,2020,
100282.
[5] Ahmadi Mehr R, Ameri A. Skin CancerDetectionBased on
Deep Learning. J Biomed Phys Eng. 2022 Dec 1;12(6):559-
568. doi: 10.31661/jbpe.v0i0.2207-1517. PMID: 36569567;
PMCID: PMC9759648.
[6] Gouda W, Sama NU, Al-Waakid G, Humayun M, Jhanjhi
NZ. Detection of Skin Cancer Based on Skin Lesion Images
Using Deep Learning. Healthcare (Basel). 2022 Jun
24;10(7):1183. doi: 10.3390/healthcare10071183. PMID:
35885710; PMCID: PMC9324455.

[7] Alwakid, G.; Gouda, W.; Humayun, M.; Sama, N.U.
Melanoma Detection Using Deep Learning-Based
Classifications. Healthcare 2022, 10, 2481.
[8] Xinrong Lu, Y. A. Firoozeh Abolhasani Zadeh, "Deep
Learning-Based ClassificationforMelanoma DetectionUsing
XceptionNet", Journal of Healthcare Engineering, vol. 2022,
Article ID 2196096, 10 pages, 2022.
[9] Ragab M, Choudhry H, Al-Rabia MW, Binyamin SS,
Aldarmahi AA and Mansour RF (2022), Early and accurate
detection of melanoma skin cancer using hybrid level set
approach. Front. Physiol. 13:965630. doi:
10.3389/fphys.2022.9656.
[10] Nawaz M, Mehmood Z, Nazir T, Naqvi RA, Rehman A,
Iqbal M, Saba T. Skin cancer detection from dermoscopic
images using deep learning and fuzzy k-means clustering.
Microsc Res Tech. 2022 Jan;85(1):339-351. doi:
10.1002/jemt.23908. Epub 2021 Aug 27. PMID: 34448519.
[11] Wu Y, Chen B, Zeng A, Pan D, Wang R and Zhao S (2022)
Skin Cancer Classification With Deep Learning:ASystematic
Review. Front. Oncol.
[12] Olusola Oluwakemi ABAYOMI-ALLI1, Robertas
DAMAŠEVIČIUS1,∗, Sanjay MISRA2, Rytis MASKELIŪNAS3,
Adebayo ABAYOMI-ALLI4Turkish Journal of Electrical
Engineering & Computer Sciences(2021).
[13] P. B. C. Castro, B. Krohling, A. G. C. Pacheco and R. A.
Krohling, "An app to detect melanoma using deep learning:
An approach to handle imbalanced data based on
evolutionary algorithms," 2020 International Joint
Conference on Neural Networks (IJCNN),Glasgow,UK,2020,
pp. 1-6, doi: 10.1109/IJCNN48605.2020.9207552.
[14] Takiddin A, Schneider J, Yang Y, Abd-Alrazaq A, Househ
M. Artificial Intelligence for Skin Cancer Detection: Scoping
Review. J Med Internet Res. 2021 Nov 24;23(11):e22934.
doi: 10.2196/22934. PMID: 34821566; PMCID:
PMC8663507.
[15] Jinnai, S.; Yamazaki, N.; Hirano, Y.; Sugawara, Y.; Ohe, Y.;
Hamamoto, R. The Development of a Skin Cancer
Classification System for PigmentedSkinLesionsUsingDeep
Learning. Biomolecules 2020, 10, 1123.
[16] Samiei, M., Hassani, A., Sarspy, S. et al. Classification of
skin cancer stages using a AHP fuzzy technique within the
context of big data healthcare. J Cancer Res Clin Oncol 149,
8743–8757 (2023).
[17] A. Mankawade, A. Bodhankar, A. Mahajan, D. Prasad, S.
Mahajan and R. Dhakalkar, "Skin Cancer Detection and
Intensity Calculator using Deep Learning," 2023
International Conference for Advancement in Technology
(ICONAT), Goa, India, 2023.
[18] Arjun, K. P. and KP Sampath Kumar. “A combined
approach of VGG 16 and LSTM transfer learning technique
for skin melanoma classification.” International journal of
health sciences (2022).
[19] Rajeshwari, J. and M. Sughasiny. “Modified PNN
classifier for diagnosing skin cancer severityconditionusing
SMO optimization technique.” AIMS Electronics and
Electrical Engineering (2022).
[20] Caroline Angelina Sunarya, Jocelyn Verna Siswanto,
Grace Shirley Cam and Felix Indra Kurniadi, “Skin Cancer
Classification using Delaunay Triangulation and Graph
Convolutional Network” International Journal of Advanced
Computer Science and Applications(IJACSA), 14(6), 2023.
[21] B. Kalpana, A.K. Reshmy, S. Senthil Pandi, S.
Dhanasekaran, OESV-KRF:Optimal ensemblesupportvector
kernel random forest basedearlydetectionandclassification
of skin diseases,Biomedical Signal Processing and
Control,Volume 85,2023,104779,ISSN
17468094,https://doi.org/10.1016/j.bspc.2023.104779.

SKIN CANCER DETECTION AND SEVERITY PREDICTION USING DEEP LEARNING

More Related Content

Similar to SKIN CANCER DETECTION AND SEVERITY PREDICTION USING DEEP LEARNING (20)

More from IRJET Journal (20)

Recently uploaded (20)

SKIN CANCER DETECTION AND SEVERITY PREDICTION USING DEEP LEARNING