Sensors,
Год журнала:
2022,
Номер
22(24), С. 9875 - 9875
Опубликована: Дек. 15, 2022
Of
the
various
tumour
types,
colorectal
cancer
and
brain
tumours
are
still
considered
among
most
serious
deadly
diseases
in
world.
Therefore,
many
researchers
interested
improving
accuracy
reliability
of
diagnostic
medical
machine
learning
models.
In
computer-aided
diagnosis,
self-supervised
has
been
proven
to
be
an
effective
solution
when
dealing
with
datasets
insufficient
data
annotations.
However,
image
often
suffer
from
irregularities,
making
recognition
task
even
more
challenging.
The
class
decomposition
approach
provided
a
robust
such
challenging
problem
by
simplifying
boundaries
dataset.
this
paper,
we
propose
model,
called
XDecompo,
improve
transferability
features
pretext
downstream
task.
XDecompo
designed
based
on
affinity
propagation-based
effectively
encourage
explainable
component
highlight
important
pixels
that
contribute
classification
explain
effect
speciality
extracted
features.
We
also
explore
generalisability
handling
different
datasets,
as
histopathology
for
images.
quantitative
results
demonstrate
robustness
high
96.16%
94.30%
CRC
images,
respectively.
demonstrated
its
generalization
capability
achieved
(both
quantitatively
qualitatively)
compared
other
Moreover,
post
hoc
method
used
validate
feature
transferability,
demonstrating
highly
accurate
representations.
Sensors,
Год журнала:
2025,
Номер
25(5), С. 1350 - 1350
Опубликована: Фев. 22, 2025
Feature
descriptors
in
histopathological
images
are
an
important
challenge
for
the
implementation
of
Content-Based
Image
Retrieval
(CBIR)
systems,
essential
tool
to
support
pathologists.
Deep
learning
models
like
Convolutional
Neural
Networks
and
Vision
Transformers
improve
extraction
these
feature
descriptors.
These
typically
generate
embeddings
by
leveraging
deeper
single-scale
linear
layers
or
advanced
pooling
layers.
However,
embeddings,
focusing
on
local
spatial
details
at
a
single
scale,
miss
out
richer
context
from
earlier
This
gap
suggests
development
methods
that
incorporate
multi-scale
information
enhance
depth
utility
image
analysis.
In
this
work,
we
propose
Local–Global
Fusion
Embedding
Model.
proposal
is
composed
three
elements:
(1)
pre-trained
backbone
multi-scales,
(2)
neck
branch
local–global
fusion,
(3)
Generalized
Mean
(GeM)-based
head
Based
our
experiments,
model’s
were
trained
ImageNet-1k
PanNuke
datasets
employing
Sub-center
ArcFace
loss
compared
with
state-of-the-art
Kimia
Path24C
dataset
retrieval,
achieving
Recall@1
99.40%
test
patches.
Senologie - Zeitschrift für Mammadiagnostik und -therapie,
Год журнала:
2025,
Номер
22(01), С. 31 - 42
Опубликована: Март 1, 2025
Abstract
Breast
pathology
poses
a
particular
diagnostic
challenge
due
to
the
broad
spectrum
of
functional,
reactive
and
neoplastic
changes
in
breast.
Objectifiable
reproducible
criteria
are
key
valid
diagnosis.
In
addition
classification
lesions,
it
is
task
pathologists
identify
document
all
tumor
characteristics
that
relevant
for
clinical
management.
Modern
personalized
medicine
based
on
up-to-date,
pathomorphological
molecular
diagnostics.
Reports
findings
should
be
written
comprehensibly,
completely
quickly.
Structured
reports
ideal
this
purpose.
Before
artificial
intelligence
can
fulfil
hopes
placed
regarding
acceleration
objectification
reporting,
technical
financial
limitations
must
resolved
explainability
AI-generated
decisions.
Mehmet Akif Ersoy Üniversitesi Uygulamalı Bilimler Dergisi,
Год журнала:
2025,
Номер
9(1), С. 63 - 74
Опубликована: Март 27, 2025
Meme
kanseri
birçok
ülkede
kadınlar
arasında
en
sık
görülen
kanser
türüdür.
kanserinin
tanı
ve
tedavisinde
verilerin
analizi
büyük
bir
önem
taşımaktadır.
Histopatolojik
görüntülerdeki
kanserli
hücre
çekirdeklerinin
segmentasyonu,
uzmanlar
için
oldukça
maliyetli
zorlu
iştir.
Bu
çalışmada,
histopatolojik
meme
görüntülerinin
çekirdek
segmentasyonu
topluluk
öğrenmesine
dayalı
LinkNet
modeli
önerilmektedir.
Görüntüler,
Kontrast
Sınırlı
Adaptif
Histogram
Eşitleme
(CLAHE)
tekniği
ile
işlendikten
sonra
veri
artırma
uygulanır.
modelinin
kodlayıcı
kısmında
ResNeXT50
Vgg19
modellerinin
yerleştirildiği
iki
ayrı
model
eğitilir.
Sonrasında,
bu
modeller
öğrenmesi
birleştirilir
maske
tahmini
yapılır.
Çalışmada
elde
edilen
0.702
Kümülatif
Jaccard
İndeks
(AJI)
metriği
sonucu,
aynı
seti
yapılmış
son
çalışmalardan
daha
başarılı
bulunmuştur.
CytoJournal,
Год журнала:
2025,
Номер
22, С. 45 - 45
Опубликована: Апрель 19, 2025
The
application
of
artificial
intelligence
(AI)
in
cancer
pathology
has
shown
significant
potential
to
enhance
diagnostic
accuracy,
streamline
workflows,
and
support
precision
oncology.
This
review
examines
the
current
applications
AI
across
various
types,
including
breast,
lung,
prostate,
colorectal
cancer,
where
aids
tissue
classification,
mutation
detection,
prognostic
predictions.
key
technologies
driving
these
advancements
include
machine
learning,
deep
computer
vision,
which
enable
automated
analysis
histopathological
images
multi-modal
data
integration.
Despite
promising
developments,
challenges
persist,
ensuring
privacy,
improving
model
interpretability,
meeting
regulatory
standards.
Furthermore,
this
explores
future
directions
AI-driven
pathology,
real-time
diagnostics,
explainable
AI,
global
accessibility,
emphasizing
importance
collaboration
between
pathologists.
Addressing
leveraging
AI's
full
could
lead
a
more
efficient,
equitable,
personalized
approach
care.
This
paper
examines
the
potential
of
Human-Centered
AI
(HCAI)
solutions
to
support
radiologists
in
diagnosing
prostate
cancer.
Prostate
cancer
is
one
most
prevalent
and
increasing
cancers
among
men.
The
scarcity
raises
concerns
about
their
ability
address
growing
demand
for
diagnosis,
leading
a
significant
surge
workload
radiologists.
Drawing
on
an
HCAI
approach,
we
sought
understand
current
practices
concerning
radiologists'
work
detecting
cancer,
as
well
challenges
they
face.
findings
from
our
empirical
studies
point
toward
that
has
expedite
informed
decision-making
enhance
accuracy,
efficiency,
consistency.
particularly
beneficial
collaborative
diagnosis
processes.
We
discuss
these
results
introduce
design
recommendations
concepts
domain
with
aim
amplifying
professional
capabilities
IEEE Access,
Год журнала:
2024,
Номер
12, С. 37557 - 37571
Опубликована: Янв. 1, 2024
Viral
and
non-viral
hepatocellular
carcinoma
(HCC)
is
becoming
predominant
in
developing
countries.
A
major
issue
linked
to
HCC-related
mortality
rate
the
late
diagnosis
of
cancer
development.
Although
traditional
approaches
diagnosing
HCC
have
become
gold-standard,
there
remain
several
limitations
due
which
confirmation
progression
takes
a
longer
period.
The
recent
emergence
artificial
intelligence
tools
with
capacity
analyze
biomedical
datasets
assisting
diagnostic
for
early
certainty.
Here
we
present
review
versus
use
(Machine
Learning
Deep
Learning)
diagnosis.
overview
cancer-related
databases
along
AI
histopathology,
radiology,
biomarker,
electronic
health
records
(EHRs)
based
given.
Oral Science International,
Год журнала:
2024,
Номер
22(1)
Опубликована: Май 19, 2024
Abstract
Purpose
This
study
aimed
to
assess
the
performance
of
ResNet‐50
deep
learning
algorithm
in
classifying
panoramic
images
determine
contact
between
mandibular
third
molars
and
inferior
alveolar
nerve
(IAN),
comparing
its
with
newly
graduated
dentists
oral
maxillofacial
surgery
specialists.
Methods
Panoramic
radiographs
were
retrieved
from
Radiology
Department,
School
Dentistry,
Kashan
University
Medical
Sciences.
The
independently
classified
as
“contact”
or
“none‐contact”
by
model,
three
dentists,
A
radiologist
sets
gold
standard
using
cone
beam‐CT.
accuracy,
precision,
recall,
specificity,
dice
coefficient
calculated
for
each
group,
inter‐rater
reliability
assessed
Cohen's
kappa
value.
Results
Of
548
retrieved,
15%
allocated
testing
dataset,
amounting
82
images.
model
showed
highest
metrics,
an
accuracy
87.80%,
precision
78.57%,
recall
84.61%,
specificity
89.28%,
81.48%.
Conversely,
novice
had
lowest
metrics
(accuracy:
74.39%
±
2.99%,
precision:
57.75%
4.33%,
recall:
74.36%
1.81%,
specificity:
74.4%
4.46%,
coefficient:
64.87%
2.69%).
Specialists
demonstrated
84.96%
1.52%,
72.65%
2.78%,
84.61%
3.14%,
85.12%
2.23%,
78.11%
1.99%.
Conclusion
Deep
algorithms
can
achieve
comparable
outcomes
specialists
may
outperform
clinicians
diagnosing
IAN.