Exploring the analysis capabilities and clinical application potential of the Claude 3 Opus in different dermatologic images: the development of a large-scale multimodal model to assist in dermatology clinical practice (Preprint)
Опубликована: Июнь 12, 2024
BACKGROUND
Publicly
available,
accessible,
and
user-friendly
artificial
intelligence
is
expected
to
serve
in
medical
processes.
Claude3,
a
newly
introduced
large-scale
multimodal
model,
has
demonstrated
significantly
superior
image
analysis
capabilities
compared
other
models
official
tests.
However,
there
no
research
reporting
the
potential
of
Claude3
analysis.
OBJECTIVE
To
explore
applications
Opus
on
dermatologic
images.
METHODS
Dermoscopy
dermatopathology
images
from
textbooks
were
collected,
question
templates
for
different
types
diseases
designed
Opus.
Three
dermatologists
used
structured
scoring
system
with
five
modules
evaluate
Opus'
based
recognition,
description,
completeness,
diagnosis,
clinical
application,
each
module
scored
out
5
total
score
25.
RESULTS
A
330
collected.
highest
pigmented
disorders
dermoscopy
(18.65/25),
followed
by
vascular
(15.97/25)
(15.86/25).
In
disorders,
its
(18.65/25)
was
higher
than
(14.54/25),
but
such
difference
existed
disorders.
Among
modules,
recognition
(3.65/5)
four
modules.
There
between
description
(3.14/5)
completeness
(3.22/5),
they
diagnostic
(2.47/5).
malignant
benign
diseases,
regardless
or
(all
p-values
<0.05),
impact
magnifications
(p>0.05)
number
evaluators.
CONCLUSIONS
exhibits
strong
disease
images,
can
accurately
describe
abnormalities
completely,
shows
high
sensitivity
diseases.
Apart
assistance,
could
potentially
be
widely
education
patient
communication.
CLINICALTRIAL
need
Язык: Английский