IEEE Access,
Journal Year:
2024,
Volume and Issue:
12, P. 67880 - 67888
Published: Jan. 1, 2024
Skin
cancer
recognition
poses
a
significant
challenge
in
the
field
of
deep
learning.
While
conventional
convolutional
neural
networks
have
been
extensively
employed
for
classifying
skin
images,
their
fixed
receptive
limits
ability
to
capture
global
features
present
such
images.
Conversely,
transformer-based
models
that
rely
on
self-attention
can
effectively
model
long-range
dependencies,
but
they
come
with
high
computational
complexity
and
exhibit
certain
limitations
local
feature
induction.
To
address
this
issue,
paper
presents
novel
network
named
ECRNet.
ECRNet
has
designed
both
information,
it
introduces
an
explicit
vision
center
accomplish
purpose.
Moreover,
fusion
module
known
as
CCPA
block.
This
utilizes
coordinate
attention
channel
mechanisms
extract
image
enhance
representation
evaluate
performance
ECRNet,
extensive
experimental
comparisons
were
conducted
ISIC2018
dataset.
The
results
demonstrate
outperforms
baseline
model,
showing
improvements
1.19%
accuracy
(ACC),
1.96%
precision,
4.08%
recall,
3.28%
F1
score.
Cogent Engineering,
Journal Year:
2024,
Volume and Issue:
11(1)
Published: Aug. 28, 2024
Skin
cancer,
a
life-threatening
disease,
necessitates
early
detection
and
accurate
classification
for
successful
treatment.
Misdiagnoses
can
lead
to
significant
consequences
patients,
highlighting
the
critical
need
improved
accuracy.
Despite
advancements
in
machine
learning
(ML)
deep
(DL)
techniques,
diagnosis
remains
challenging
due
its
complex
nature.
This
comprehensive
Systematic
Literature
Review
(SLR)
aims
examine
use
of
ML
DL
models
skin
cancer
classification,
including
convolutional
neural
networks
(CNNs),
recurrent
(RNNs),
transfer
learning.
Model
performance
is
evaluated
based
on
accuracy,
sensitivity,
specificity,
precision.
Key
findings
reveal
dominance
models,
with
SVM-PSO
emerging
as
top-performing
hybrid
model
97.50%
Tailored
such
M-SVM
FCN-ResAlexNet,
demonstrate
high
emphasizing
importance
customization
dermatology
tasks.
Deep
networks,
ResNet-50,
ResNet34,
Inception
V3,
ResNet
152,
consistently
exhibit
strong
performance,
impact
architectural
depth.
Traditional
algorithms
like
Random
Forest,
KNN,
Naive
Bayes
face
challenges
compared
models.
Furthermore,
analysis
explores
correlations
between
dataset
size
revealing
varied
responses.
Temporal
trends
model-specific
analyses
uncover
outliers,
anomalies,
influence
specific
datasets
(e.g.
imbalanced
classes),
providing
valuable
insights
future
research
development.
multifaceted
nature
influenced
by
factors
beyond
size,
underscores
nuanced
considerations
image
classification.
Overall,
this
SLR
offer
researchers
practitioners,
serving
crucial
step
towards
developing
even
more
reliable
tools
diagnosis.
Automatika,
Journal Year:
2024,
Volume and Issue:
65(4), P. 1458 - 1468
Published: Aug. 29, 2024
Skin
cancer
has
become
a
serious
disease
which
the
potential
to
scale
up
if
it
is
not
identified
earlier.
It
imperative
detect
and
give
treatment
skin
promptly.
Diagnosing
manually
takes
lot
of
time
costly,
probability
false
diagnosis
increased
due
outstanding
resemblances
among
various
lesions.
Enhancing
classification
multi-class
lesions
needs
development
investigative
systems
should
be
automated.
Data
augmentation
with
GANs
Adaptive
Conditional
Generative
Adversarial
Network
strategies
improves
performance.
The
performance
tested
using
balanced
unbalanced
datasets.
Using
proper
process
data,
suggested
system
attains
94%
accuracy
for
VGG16,
93%
ResNet50
94.25%
ResNet101.
collaboration
all
such
methods
further
95%.
In
summary,
novelty
work
lies
in
its
holistic
approach
automated
lesion
classification,
incorporating
advanced
deep
learning
models,
novel
data
techniques
comprehensive
evaluation
on
real-world
These
contributions
collectively
advance
field
computer-aided
detection
treatment.
