medRxiv (Cold Spring Harbor Laboratory),
Год журнала:
2023,
Номер
unknown
Опубликована: Авг. 9, 2023
Abstract
Background
Large
Language
Models
(LLMs)
like
GPT-4
demonstrate
potential
applications
in
diverse
areas,
including
healthcare
and
patient
education.
This
study
evaluates
GPT-4’s
competency
against
osteoarthritis
(OA)
treatment
guidelines
from
the
United
States
China
assesses
its
ability
diagnosing
treating
orthopedic
diseases.
Methods
Data
sources
included
OA
management
examination
case
questions.
Queries
were
directed
to
based
on
these
resources,
responses
compared
with
established
cases.
The
accuracy
completeness
of
evaluated
using
Likert
scales,
while
inquiries
stratified
into
four
tiers
correctness
completeness.
Results
exhibited
strong
performance
providing
accurate
complete
recommendations
both
American
Chinese
guidelines,
high
scale
scores
for
It
demonstrated
proficiency
handling
clinical
cases,
making
diagnoses,
suggesting
appropriate
tests,
proposing
plans.
Few
errors
noted
specific
complex
Conclusions
exhibits
as
an
auxiliary
tool
practice
education,
demonstrating
interpreting
analyzing
Further
validation
capabilities
real-world
scenarios
is
needed.
Scientific Reports,
Год журнала:
2023,
Номер
13(1)
Опубликована: Ноя. 22, 2023
Abstract
The
study
aimed
to
evaluate
the
performance
of
two
Large
Language
Models
(LLMs):
ChatGPT
(based
on
GPT-3.5)
and
GPT-4
with
temperature
parameter
values,
Polish
Medical
Final
Examination
(MFE).
models
were
tested
three
editions
MFE
from:
Spring
2022,
Autumn
2023
in
language
versions—English
Polish.
accuracies
both
compared
relationships
between
correctness
answers
answer’s
metrics
investigated.
demonstrated
that
outperformed
GPT-3.5
all
examinations
regardless
used.
achieved
mean
79.7%
for
English
versions,
passing
versions.
had
54.8%
60.3%
English,
none
2
3
versions
equal
0
1
respectively
while
value.
score
was
mostly
lower
than
average
a
medical
student.
There
statistically
significant
correlation
index
difficulty
models.
overall
accuracy
still
suboptimal
worse
students.
This
emphasizes
need
further
improvements
LLMs
before
they
can
be
reliably
deployed
settings.
These
findings
suggest
an
increasing
potential
usage
terms
education.
Bone and Joint Research,
Год журнала:
2023,
Номер
12(7), С. 447 - 454
Опубликована: Июль 10, 2023
The
use
of
artificial
intelligence
(AI)
is
rapidly
growing
across
many
domains,
which
the
medical
field
no
exception.
AI
an
umbrella
term
defining
practical
application
algorithms
to
generate
useful
output,
without
need
human
cognition.
Owing
expanding
volume
patient
information
collected,
known
as
'big
data',
showing
promise
a
tool
in
healthcare
research
and
all
aspects
care
pathways.
Practical
applications
orthopaedic
surgery
include:
diagnostics,
such
fracture
recognition
tumour
detection;
predictive
models
clinical
patient-reported
outcome
measures,
calculating
mortality
rates
length
hospital
stay;
real-time
rehabilitation
monitoring
surgical
training.
However,
clinicians
should
remain
cognizant
AI's
limitations,
development
robust
reporting
validation
frameworks
paramount
importance
prevent
avoidable
errors
biases.
aim
this
review
article
provide
comprehensive
understanding
its
subfields,
well
delineate
existing
trauma
surgery.
Furthermore,
narrative
expands
upon
limitations
future
direction.
As
artificial
intelligence
(AI)
continues
to
evolve
and
mature,
it
is
increasingly
finding
applications
in
the
field
of
healthcare,
particularly
specialties
like
radiology
that
are
data-heavy
image-focused.
Language
learning
models
(LLMs)
such
as
OpenAI's
Generative
Pre-trained
Transformer-4
(GPT-4)
new
medicine
there
a
paucity
literature
regarding
possible
utilities
GPT-4
given
its
novelty.
We
aim
present
an
in-depth
exploration
role
GPT-4,
advanced
language
model,
radiology.
Giving
model
prompts
for
generating
reports,
template
generation,
enhancing
clinical
decision-making,
suggesting
captivating
titles
research
articles,
patient
communication,
education,
can
occasionally
be
quite
generic,
at
times,
may
factually
incorrect
content,
which
could
lead
errors.
The
responses
were
then
analyzed
detail
their
potential
utility
day-to-day
radiologist
workflow,
processes.
Further
required
evaluate
LLMs'
accuracy
safety
practice
develop
comprehensive
guidelines
implementation.
Journal of Human Sport and Exercise,
Год журнала:
2024,
Номер
20(1), С. 39 - 56
Опубликована: Авг. 6, 2024
The
study
aims
to
investigate
the
effect
of
a
5-week
artificial
intelligence-generated
calisthenics
training
program
(AIGCTP)
on
health-related
physical
fitness
components,
including
flexibility,
cardiovascular
endurance,
and
muscular
endurance.
Utilizing
quasi-experimental
design,
employed
one-group
pre-test-post-test
design
for
within-group
comparisons
two-group
between-group
comparisons.
Participants
included
87
untrained
collegiate
students,
divided
into
AIGCTP
group
(43
participants)
human-made
(HMCTP)
(44
participants),
selected
via
purposive
sampling.
A
paired
t-test
was
used
comparisons,
an
independent
sample
findings
indicated
that
effectively
improved
flexibility
lower
extremities
endurance
core
upper
extremities.
However,
female
participants
did
not
show
significant
improvements
in
any
whereas
male
demonstrated
HMCTP
effective
improving
all
participants.
Between-group
revealed
significantly
superior
group,
irrespective
sex.
Additionally,
males
exhibited
higher
compared
those
group.
suggests
AI
can
be
training,
but
professional-made
programs
are
some
areas.
Future
research
should
replicate
these
findings,
examine
more
explore
longer
durations
further
validation.
SLAS TECHNOLOGY,
Год журнала:
2024,
Номер
29(4), С. 100162 - 100162
Опубликована: Июль 4, 2024
This
study
presents
a
scientometric
analysis
of
the
intersection
between
rehabilitation
science
and
artificial
intelligence
(AI)
technologies,
using
data
from
Web
Science
(WOS)
database
2002
to
2022.
The
employed
comprehensive
search
query
with
key
AI-related
terms,
focusing
on
wide
range
publications
in
science.
Utilizing
Citespace
tool,
visualizes
quantifies
relationships
identifies
research
trends,
assesses
impact
AI
technologies
Findings
reveal
significant
increase
this
field,
particularly
2017
onwards,
peaking
2021.
United
States
has
been
leading
contributor,
followed
by
countries
like
England,
Australia,
Germany,
Canada.
Major
institutional
contributions
come
Harvard
University
Pennsylvania
Commonwealth
System
Higher
Education,
among
others.
A
keyword
co-occurrence
network
constructed
through
nine
distinct
hot
topics
various
frontiers,
highlighting
evolving
focus
areas
within
field.
Burst
keywords
indicates
shift
performance
injury-related
an
increasing
emphasis
deep
learning
recent
years.
also
predicts
potential
papers,
spotlighting
works
Kunze
KN
others
as
significantly
influencing
future
directions.
Additionally,
it
examines
evolution
knowledge
bases
research,
revealing
multidisciplinary
core
that
includes
neurology,
rehabilitation,
ophthalmology,
extending
complementary
fields
such
medicine
social
sciences.
provides
overview
AI's
application
science,
offering
insights
into
its
evolution,
impact,
emerging
trends
over
past
two
decades.
findings
suggest
strategic
directions
for
policy-making,
interdisciplinary
collaboration
AI.
medRxiv (Cold Spring Harbor Laboratory),
Год журнала:
2023,
Номер
unknown
Опубликована: Июнь 5, 2023
Abstract
Introduction
The
rapid
progress
in
artificial
intelligence,
machine
learning,
and
natural
language
processing
has
led
to
the
emergence
of
increasingly
sophisticated
large
models
(LLMs)
enabling
their
use
various
applications,
including
medicine
healthcare.
Objectives
study
aimed
evaluate
performance
two
LLMs:
ChatGPT
(based
on
GPT-3.5)
GPT-4,
Medical
Final
Examination
(MFE).
Methods
were
tested
three
editions
MFE
from:
Spring
2022,
Autumn
2023
versions
–
English
Polish.
accuracies
both
compared
relationships
between
correctness
answers
with
index
difficulty
discrimination
power
investigated.
Results
demonstrated
that
GPT-4
outperformed
GPT-3.5
all
examinations
regardless
used.
achieved
mean
80.7%
for
Polish
79.6%
English,
passing
versions.
had
56.6%
58.3%
2
3
test.
score
was
lower
than
average
a
medical
student.
There
significant
positive
negative
correlation
index,
respectively,
exams.
Conclusions
These
findings
contribute
growing
body
literature
utility
LLMs
medicine.
They
also
suggest
an
increasing
potential
usage
terms
education
decision-making
support.
What’s
new?
Recent
advancements
intelligence
have
resulted
development
(LLMs).
This
focused
evaluation
LLMs,
across
from
editions.
study,
best
our
knowledge,
presents
first
validation
those
European-based
final
examinations.
exams,
achieving
accuracy
(Polish)
(English),
while
attained
(English)
respectively.
However,
GPT-4’s
scores
fell
short
typical
student
performance.
understanding
LLM’s
hint
at
