Frontiers in Medicine,
Journal Year:
2024,
Volume and Issue:
11
Published: Aug. 22, 2024
Introduction
Nodal
metastasis
(NM)
in
sentinel
node
biopsies
(SNB)
is
crucial
for
melanoma
staging.
However,
an
intra-nodal
nevus
(INN)
may
often
be
misclassified
as
NM,
leading
to
potential
misdiagnosis
and
incorrect
There
high
discordance
among
pathologists
assessing
SNB
positivity,
which
lead
false
Digital
whole
slide
imaging
offers
the
implementing
artificial
intelligence
(AI)
digital
pathology.
In
this
study,
we
assessed
capability
of
AI
detect
NM
INN
SNBs.
Methods
A
total
485
hematoxylin
eosin
images
(WSIs),
including
from
196
SNBs,
were
collected
divided
into
training
(279
WSIs),
validation
(89
test
sets
(117
WSIs).
deep
learning
model
was
trained
with
5,956
manual
pixel-wise
annotations.
The
three
blinded
dermatopathologists
set,
immunohistochemistry
serving
reference
standard.
Results
showed
excellent
performance
area
under
curve
receiver
operating
characteristic
(AUC)
0.965
detecting
NM.
comparison,
AUC
detection
ranged
between
0.94
0.98.
For
INN,
lower
both
(0.781)
(range
0.63–0.79).
Discussion
conclusion,
accuracy
achieving
dermatopathologist-level
INN.
Importantly,
differentiate
these
two
entities.
further
warranted.
Journal of Clinical Pathology,
Journal Year:
2024,
Volume and Issue:
unknown, P. jcp - 209304
Published: Jan. 10, 2024
Aims
To
evaluate
the
accuracy
of
Chat
Generative
Pre-trained
Transformer
(ChatGPT)
powered
by
GPT-4
in
histopathological
image
detection
and
classification
colorectal
adenomas
using
diagnostic
consensus
provided
pathologists
as
a
reference
standard.
Methods
A
study
was
conducted
with
100
polyp
photomicrographs,
comprising
an
equal
number
non-adenomas,
classified
two
pathologists.
These
images
were
analysed
classic
for
1
time
October
2023
custom
20
times
December
2023.
GPT-4’s
responses
compared
against
standard
through
statistical
measures
to
its
proficiency
diagnosis,
further
assessing
model’s
descriptive
accuracy.
Results
demonstrated
median
sensitivity
74%
specificity
36%
adenoma
detection.
The
varied,
ranging
from
16%
non-specific
changes
tubular
adenomas.
Its
consistency,
indicated
low
kappa
values
0.06
0.11,
suggested
only
poor
slight
agreement.
All
microscopic
descriptions
corresponded
their
diagnoses.
also
commented
about
limitations
diagnoses
(eg,
slide
diagnosis
best
done
pathologists,
inadequacy
single-image
conclusions,
need
clinical
data
higher
magnification
view).
Conclusions
showed
high
but
detecting
varied
classification.
However,
consistency
low.
This
artificial
intelligence
tool
acknowledged
limitations,
emphasising
pathologist’s
expertise
additional
context.
Hepatology Communications,
Journal Year:
2025,
Volume and Issue:
9(1)
Published: Jan. 1, 2025
Histopathologic
evaluation
of
liver
biopsy
has
played
a
longstanding
role
in
the
diagnosis
and
management
disease.
However,
utility
been
questioned
by
some,
given
improved
imaging
modalities,
increased
availability
noninvasive
serologic
tests,
development
artificial
intelligence
over
past
several
years.
In
this
review,
we
discuss
current
future
both
non-neoplastic
neoplastic
diseases
era
laboratory,
radiologic,
digital
technologies.
Annals of Internal Medicine,
Journal Year:
2025,
Volume and Issue:
unknown
Published: April 4, 2025
Whether
artificial
intelligence
(AI)
assistance
is
associated
with
quality
of
care
uncertain.
To
compare
initial
AI
recommendations
final
physicians
who
had
access
to
the
and
may
or
not
have
viewed
them.
Retrospective
cohort
study.
Cedars-Sinai
Connect,
an
AI-assisted
virtual
urgent
clinic
intake
questions
via
structured
chat.
When
confidence
sufficient,
presents
diagnosis
management
(prescriptions,
laboratory
tests,
referrals).
461
physician-managed
visits
sufficient
complete
medical
records
for
adults
respiratory,
urinary,
vaginal,
eye,
dental
symptoms
from
12
June
14
July
2024.
Concordance
physician
recommendations.
Physician
adjudicators
scored
all
nonconcordant
a
sample
concordant
as
optimal,
reasonable,
inadequate,
potentially
harmful.
Initial
were
262
(56.8%).
Among
weighted
visits,
more
frequently
rated
optimal
(77.1%
[95%
CI,
72.7%
80.9%])
compared
treating
decisions
(67.1%
[CI,
62.9%
71.1%]).
Quality
scores
equal
in
67.9%
(CI,
64.8%
70.9%)
cases,
better
20.8%
17.8%
24.0%),
11.3%
9.0%
14.2%),
respectively.
Single-center
retrospective
Adjudicators
blinded
source
It
unknown
whether
differed,
often
quality.
Findings
suggest
that
performed
identifying
critical
red
flags
supporting
guideline-adherent
care,
whereas
at
adapting
changing
information
during
consultations.
Thus,
role
assisting
decision
making
care.
K
Health.
Bioengineering,
Journal Year:
2024,
Volume and Issue:
11(4), P. 342 - 342
Published: March 31, 2024
Large
language
models
(LLMs)
are
transformer-based
neural
networks
that
can
provide
human-like
responses
to
questions
and
instructions.
LLMs
generate
educational
material,
summarize
text,
extract
structured
data
from
free
create
reports,
write
programs,
potentially
assist
in
case
sign-out.
combined
with
vision
interpreting
histopathology
images.
have
immense
potential
transforming
pathology
practice
education,
but
these
not
infallible,
so
any
artificial
intelligence
generated
content
must
be
verified
reputable
sources.
Caution
exercised
on
how
integrated
into
clinical
practice,
as
produce
hallucinations
incorrect
results,
an
over-reliance
may
lead
de-skilling
automation
bias.
This
review
paper
provides
a
brief
history
of
highlights
several
use
cases
for
the
field
pathology.
