JMIR AI,
Год журнала:
2024,
Номер
4, С. e63701 - e63701
Опубликована: Дек. 2, 2024
Background
Chat-based
counseling
services
are
popular
for
the
low-threshold
provision
of
mental
health
support
to
youth.
In
addition,
they
particularly
suitable
utilization
natural
language
processing
(NLP)
improved
care.
Objective
Consequently,
this
paper
evaluates
feasibility
such
a
use
case,
namely,
NLP-based
automated
evaluation
satisfaction
with
chat
interaction.
This
preregistered
approach
could
be
used
and
quality
control
procedures,
as
it
is
relevant
those
services.
Methods
The
consultations
2609
young
chatters
(around
140,000
messages)
corresponding
feedback
were
train
evaluate
classifiers
predict
whether
was
perceived
helpful
or
not.
On
one
hand,
we
trained
word
vectorizer
in
combination
an
extreme
gradient
boosting
(XGBoost)
classifier,
applying
cross-validation
extensive
hyperparameter
tuning.
other
several
transformer-based
models,
comparing
model
types,
preprocessing,
over-
undersampling
techniques.
For
both
selected
best-performing
on
training
set
final
performance
522
users
test
set.
Results
fine-tuned
XGBoost
classifier
achieved
area
under
receiver
operating
characteristic
score
0.69
(P<.001),
well
Matthews
correlation
coefficient
0.25
previously
unseen
Longformer-based
did
not
outperform
baseline,
scoring
0.68
(P=.69).
A
Shapley
additive
explanations
explainability
suggested
that
help
seekers
rating
consultation
commonly
expressed
their
already
within
conversation.
contrast,
rejection
offered
exercises
predicted
unhelpfulness.
Conclusions
Chat
conversations
include
information
regarding
interaction
can
by
prediction
approaches.
However,
determine
if
moderate
predictive
translates
into
meaningful
service
improvements
requires
randomized
trials.
Further,
our
results
highlight
relevance
contrasting
pretrained
models
simpler
baselines
avoid
implementation
unnecessarily
complex
models.
Trial
Registration
Open
Science
Framework
SR4Q9;
https://osf.io/sr4q9
Bioengineering,
Год журнала:
2024,
Номер
11(4), С. 337 - 337
Опубликована: Март 29, 2024
As
healthcare
systems
around
the
world
face
challenges
such
as
escalating
costs,
limited
access,
and
growing
demand
for
personalized
care,
artificial
intelligence
(AI)
is
emerging
a
key
force
transformation.
This
review
motivated
by
urgent
need
to
harness
AI’s
potential
mitigate
these
issues
aims
critically
assess
integration
in
different
domains.
We
explore
how
AI
empowers
clinical
decision-making,
optimizes
hospital
operation
management,
refines
medical
image
analysis,
revolutionizes
patient
care
monitoring
through
AI-powered
wearables.
Through
several
case
studies,
we
has
transformed
specific
domains
discuss
remaining
possible
solutions.
Additionally,
will
methodologies
assessing
solutions,
ethical
of
deployment,
importance
data
privacy
bias
mitigation
responsible
technology
use.
By
presenting
critical
assessment
transformative
potential,
this
equips
researchers
with
deeper
understanding
current
future
impact
on
healthcare.
It
encourages
an
interdisciplinary
dialogue
between
researchers,
clinicians,
technologists
navigate
complexities
implementation,
fostering
development
AI-driven
solutions
that
prioritize
standards,
equity,
patient-centered
approach.
Abstract
Background
A
comprehensive
overview
of
artificial
intelligence
(AI)
for
cardiovascular
disease
(CVD)
prediction
and
a
screening
tool
AI
models
(AI-Ms)
independent
external
validation
are
lacking.
This
systematic
review
aims
to
identify,
describe,
appraise
AI-Ms
CVD
in
the
general
special
populations
develop
new
score
(IVS)
replicability
evaluation.
Methods
PubMed,
Web
Science,
Embase,
IEEE
library
were
searched
up
July
2021.
Data
extraction
analysis
performed
populations,
distribution,
predictors,
algorithms,
etc.
The
risk
bias
was
evaluated
with
assessment
(PROBAST).
Subsequently,
we
designed
IVS
model
evaluation
five
steps
items,
including
transparency
performance
models,
feasibility
reproduction,
clinical
implication,
respectively.
is
registered
PROSPERO
(No.
CRD42021271789).
Results
In
20,887
screened
references,
79
articles
(82.5%
2017–2021)
included,
which
contained
114
datasets
(67
Europe
North
America,
but
0
Africa).
We
identified
486
AI-Ms,
majority
development
(
n
=
380),
none
them
had
undergone
validation.
total
66
idiographic
algorithms
found;
however,
36.4%
used
only
once
39.4%
over
three
times.
large
number
different
predictors
(range
5–52,000,
median
21)
large-span
sample
size
80–3,660,000,
4466)
observed.
All
at
high
according
PROBAST,
primarily
due
incorrect
use
statistical
methods.
confirmed
10
as
“recommended”;
281
187
“not
recommended”
“warning,”
Conclusion
has
led
digital
revolution
field
prediction,
still
early
stage
defects
research
design,
report,
systems.
developed
may
contribute
this
field.
Diagnostic and Interventional Radiology,
Год журнала:
2024,
Номер
0(0), С. 0 - 0
Опубликована: Фев. 20, 2024
To
determine
how
radiology,
nuclear
medicine,
and
medical
imaging
journals
encourage
mandate
the
use
of
reporting
guidelines
for
artificial
intelligence
(AI)
in
their
author
reviewer
instructions.
METHODSThe
primary
source
journal
information
associated
citation
data
used
was
Journal
Citation
Reports
(June
2023
release
2022
data;
Clarivate
Analytics,
UK).The
first-and
second-quartile
indexed
Science
Index
Expanded
Emerging
Sources
were
included.The
instructions
evaluated
by
two
independent
readers,
followed
an
additional
reader
consensus,
with
assistance
automatic
annotation.Encouragement
submission
requirements
systematically
analyzed.The
grouped
as
AI-specific,
related
to
modeling,
unrelated
modeling.
RESULTSOut
102
journals,
98
included
this
study,
all
them
had
instructions.Only
five
(5%)
encouraged
authors
follow
AI-specific
guidelines.Among
these,
three
required
a
filled-out
checklist.Reviewer
found
16
(16%),
among
which
one
(6%)
reviewers
without
requirements.The
proportions
encouragement
statistically
significantly
lower
compared
those
other
types
(P
<
0.05
all).
CONCLUSIONThe
findings
indicate
that
are
not
commonly
mandated
(i.e.,
requiring
checklist)
these
modeling
leaving
vast
space
improvement.This
meta-research
study
hopes
contribute
awareness
community
AI
ignite
large-scale
group
efforts
stakeholders,
making
research
less
wasteful.
CLINICAL
SIGNIFICANCEThis
highlights
need
improved
journals.This
can
potentially
foster
greater
motivate
various
stakeholders
collaborate
promote
more
efficient
responsible
practices.
International Journal of Eating Disorders,
Год журнала:
2024,
Номер
57(6), С. 1357 - 1368
Опубликована: Апрель 10, 2024
To
provide
a
brief
overview
of
artificial
intelligence
(AI)
application
within
the
field
eating
disorders
(EDs)
and
propose
focused
solutions
for
research.
Journal of Medical Internet Research,
Год журнала:
2024,
Номер
26, С. e52508 - e52508
Опубликована: Май 2, 2024
The
number
of
papers
presenting
machine
learning
(ML)
models
that
are
being
submitted
to
and
published
in
the
Journal
Medical
Internet
Research
other
JMIR
Publications
journals
has
steadily
increased.
Editors
peer
reviewers
involved
review
process
for
such
manuscripts
often
go
through
multiple
cycles
enhance
quality
completeness
reporting.
use
reporting
guidelines
or
checklists
can
help
ensure
consistency
(and
published)
scientific
and,
example,
avoid
instances
missing
information.
In
this
Editorial,
editors
discuss
general
policy
regarding
authors’
application
specifically
focus
on
ML
studies
journals,
using
Consolidated
Reporting
Machine
Learning
Studies
(CREMLS)
guidelines,
with
an
example
how
authors
could
CREMLS
checklist
transparency
rigor
Journal of Medical Internet Research,
Год журнала:
2024,
Номер
26, С. e47645 - e47645
Опубликована: Июнь 13, 2024
In
recent
years,
there
has
been
explosive
development
in
artificial
intelligence
(AI),
which
widely
applied
the
health
care
field.
As
a
typical
AI
technology,
machine
learning
models
have
emerged
with
great
potential
predicting
cardiovascular
diseases
by
leveraging
large
amounts
of
medical
data
for
training
and
optimization,
are
expected
to
play
crucial
role
reducing
incidence
mortality
rates
diseases.
Although
field
become
research
hot
spot,
still
many
pitfalls
that
researchers
need
pay
close
attention
to.
These
may
affect
predictive
performance,
credibility,
reliability,
reproducibility
studied
models,
ultimately
value
affecting
prospects
clinical
application.
Therefore,
identifying
avoiding
these
is
task
before
implementing
research.
However,
currently
lack
comprehensive
summary
on
this
topic.
This
viewpoint
aims
analyze
existing
problems
terms
quality,
set
characteristics,
model
design,
statistical
methods,
as
well
implications,
provide
possible
solutions
problems,
such
gathering
objective
data,
improving
training,
repeating
measurements,
increasing
sample
size,
preventing
overfitting
using
specific
algorithms
address
targeted
issues,
standardizing
outcomes
evaluation
criteria,
enhancing
fairness
replicability,
goal
offering
reference
assistance
researchers,
algorithm
developers,
policy
makers,
practitioners.
medRxiv (Cold Spring Harbor Laboratory),
Год журнала:
2024,
Номер
unknown
Опубликована: Июль 22, 2024
Large
language
models
(LLMs)
show
promise
in
supporting
differential
diagnosis,
but
their
performance
is
challenging
to
evaluate
due
the
unstructured
nature
of
responses.
To
assess
current
capabilities
LLMs
diagnose
genetic
diseases,
we
benchmarked
these
on
5,213
case
reports
using
Phenopacket
Schema,
Human
Phenotype
Ontology
and
Mondo
disease
ontology.
Prompts
generated
from
each
phenopacket
were
sent
three
generative
pretrained
transformer
(GPT)
models.
The
same
phenopackets
used
as
input
a
widely
diagnostic
tool,
Exomiser,
phenotype-only
mode.
best
LLM
ranked
correct
diagnosis
first
23.6%
cases,
whereas
Exomiser
did
so
35.5%
cases.
While
for
has
been
improving,
it
not
reached
level
commonly
traditional
bioinformatics
tools.
Future
research
needed
determine
approach
incorporate
into
pipelines.
Journal of Medical Internet Research,
Год журнала:
2025,
Номер
27, С. e67256 - e67256
Опубликована: Янв. 7, 2025
Background
Oral
microenvironmental
disorders
are
associated
with
an
increased
risk
of
heart
failure
preserved
ejection
fraction
(HFpEF).
Hyperspectral
imaging
(HSI)
technology
enables
the
detection
substances
that
visually
indistinguishable
to
human
eye,
providing
a
noninvasive
approach
extensive
applications
in
medical
diagnostics.
Objective
The
objective
this
study
is
develop
and
validate
digital,
oral
diagnostic
model
for
patients
HFpEF
using
HSI
combined
various
machine
learning
algorithms.
Methods
Between
April
2023
August
2023,
total
140
were
recruited
from
Renmin
Hospital
Wuhan
University
serve
as
training
internal
testing
groups
study.
Subsequently,
2024
September
2024,
additional
35
enrolled
Three
Gorges
Yichang
Central
People’s
constitute
external
group.
After
preprocessing
ensure
image
quality,
spectral
textural
features
extracted
images.
We
25
bands
each
patient
obtained
8
corresponding
texture
evaluate
performance
28
algorithms
their
ability
distinguish
control
participants
HFpEF.
demonstrating
optimal
both
was
selected
construct
model.
significant
identifying
identified
further
interpretative
analysis.
Shapley
Additive
Explanations
(SHAP)
used
provide
analytical
insights
into
feature
importance.
Results
Participants
divided
group
(n=105),
(n=35),
consistent
baseline
characteristics
across
groups.
Among
tested,
random
forest
algorithm
demonstrated
superior
area
under
receiver
operating
characteristic
curve
(AUC)
0.884
accuracy
82.9%
group,
well
AUC
0.812
85.7%
For
interpretation,
we
top
by
algorithm.
SHAP
analysis
revealed
discernible
distinctions
between
HFpEF,
thereby
validating
model’s
capacity
accurately
identify
Conclusions
This
efficient
facilitates
identification
individuals
promoting
early
detection,
diagnosis,
treatment.
Our
research
presents
clinically
advanced
framework
validated
independent
data
sets
potential
enhance
care.
Trial
Registration
China
Clinical
Registry
ChiCTR2300078855;
https://www.chictr.org.cn/showproj.html?proj=207133