Frontiers of Medicine,
Год журнала:
2024,
Номер
18(5), С. 778 - 797
Опубликована: Авг. 8, 2024
Abstract
Cancer
is
a
heterogeneous
and
multifaceted
disease
with
significant
global
footprint.
Despite
substantial
technological
advancements
for
battling
cancer,
early
diagnosis
selection
of
effective
treatment
remains
challenge.
With
the
convenience
large-scale
datasets
including
multiple
levels
data,
new
bioinformatic
tools
are
needed
to
transform
this
wealth
information
into
clinically
useful
decision-support
tools.
In
field,
artificial
intelligence
(AI)
technologies
their
highly
diverse
applications
rapidly
gaining
ground.
Machine
learning
methods,
such
as
Bayesian
networks,
support
vector
machines,
decision
trees,
random
forests,
gradient
boosting,
K-nearest
neighbors,
neural
network
models
like
deep
learning,
have
proven
valuable
in
predictive,
prognostic,
diagnostic
studies.
Researchers
recently
employed
large
language
tackle
dimensions
problems.
However,
leveraging
opportunity
utilize
AI
clinical
settings
will
require
surpassing
obstacles—a
major
issue
lack
use
available
reporting
guidelines
obstructing
reproducibility
published
review,
we
discuss
methods
explore
benefits
limitations.
We
summarize
healthcare
highlight
potential
role
impact
on
future
directions
cancer
research.
Physics in Medicine and Biology,
Год журнала:
2023,
Номер
68(23), С. 23TR01 - 23TR01
Опубликована: Сен. 18, 2023
Breast
cancer,
which
is
the
most
common
type
of
malignant
tumor
among
humans,
a
leading
cause
death
in
females.
Standard
treatment
strategies,
including
neoadjuvant
chemotherapy,
surgery,
postoperative
targeted
therapy,
endocrine
and
radiotherapy,
are
tailored
for
individual
patients.
Such
personalized
therapies
have
tremendously
reduced
threat
breast
cancer
Furthermore,
early
imaging
screening
plays
an
important
role
reducing
cycle
improving
prognosis.
The
recent
innovative
revolution
artificial
intelligence
(AI)
has
aided
radiologists
accurate
diagnosis
cancer.
In
this
review,
we
introduce
necessity
incorporating
AI
into
applications
mammography,
ultrasonography,
magnetic
resonance
imaging,
positron
emission
tomography/computed
tomography
based
on
published
articles
since
1994.
Moreover,
challenges
discussed.
Abstract
Breast
cancer
holds
the
highest
diagnosis
rate
among
female
tumors
and
is
leading
cause
of
death
women.
Quantitative
analysis
radiological
images
shows
potential
to
address
several
medical
challenges,
including
early
detection
classification
breast
tumors.
In
P.I.N.K
study,
66
women
were
enrolled.
Their
paired
Automated
Volume
Scanner
(ABVS)
Digital
Tomosynthesis
(DBT)
images,
annotated
with
cancerous
lesions,
populated
first
ABVS+DBT
dataset.
This
enabled
not
only
a
radiomic
for
malignant
vs.
benign
classification,
but
also
comparison
two
modalities.
For
this
purpose,
models
trained
using
leave-one-out
nested
cross-validation
strategy
combined
proper
threshold
selection
approach.
approach
provides
statistically
significant
results
even
medium-sized
data
sets.
Additionally
it
distributional
variables
importance,
thus
identifying
most
informative
features.
The
proved
predictive
capacity
reduced
number
Indeed,
from
tomography
we
achieved
AUC-ROC
$$89.9\%$$
89.9%
19
features
$$92.1\%$$
92.1
7
them;
while
ABVS
attained
an
$$72.3\%$$
72.3
22
$$85.8\%$$
85.8
3
Although
power
DBT
outperforms
ABVS,
when
comparing
predictions
at
patient
level,
8.7%
lesions
are
misclassified
by
both
methods,
suggesting
partial
complementarity.
Notably,
promising
(AUC-ROC
ABVS-DBT
$$71.8\%$$
71.8
-
$$74.1\%$$
74.1
)
non-geometric
features,
opening
way
integration
virtual
biopsy
in
routine.
Abstract
The
advent
of
radiomics
has
revolutionized
medical
image
analysis,
affording
the
extraction
high
dimensional
quantitative
data
for
detailed
examination
normal
and
abnormal
tissues.
Artificial
intelligence
(AI)
can
be
used
enhancement
a
series
steps
in
pipeline,
from
acquisition
preprocessing,
to
segmentation,
feature
extraction,
selection,
model
development.
aim
this
review
is
present
most
AI
methods
explaining
advantages
limitations
methods.
Some
prominent
architectures
mentioned
include
Boruta,
random
forests,
gradient
boosting,
generative
adversarial
networks,
convolutional
neural
transformers.
Employing
these
models
process
analysis
significantly
enhance
quality
effectiveness
while
addressing
several
that
reduce
predictions.
Addressing
enable
clinical
decisions
wider
adoption.
Importantly,
will
highlight
how
assist
overcoming
major
bottlenecks
implementation,
ultimately
improving
translation
potential
method.
Frontiers of Medicine,
Год журнала:
2024,
Номер
18(5), С. 778 - 797
Опубликована: Авг. 8, 2024
Abstract
Cancer
is
a
heterogeneous
and
multifaceted
disease
with
significant
global
footprint.
Despite
substantial
technological
advancements
for
battling
cancer,
early
diagnosis
selection
of
effective
treatment
remains
challenge.
With
the
convenience
large-scale
datasets
including
multiple
levels
data,
new
bioinformatic
tools
are
needed
to
transform
this
wealth
information
into
clinically
useful
decision-support
tools.
In
field,
artificial
intelligence
(AI)
technologies
their
highly
diverse
applications
rapidly
gaining
ground.
Machine
learning
methods,
such
as
Bayesian
networks,
support
vector
machines,
decision
trees,
random
forests,
gradient
boosting,
K-nearest
neighbors,
neural
network
models
like
deep
learning,
have
proven
valuable
in
predictive,
prognostic,
diagnostic
studies.
Researchers
recently
employed
large
language
tackle
dimensions
problems.
However,
leveraging
opportunity
utilize
AI
clinical
settings
will
require
surpassing
obstacles—a
major
issue
lack
use
available
reporting
guidelines
obstructing
reproducibility
published
review,
we
discuss
methods
explore
benefits
limitations.
We
summarize
healthcare
highlight
potential
role
impact
on
future
directions
cancer
research.
