Briefings in Bioinformatics,
Год журнала:
2024,
Номер
26(1)
Опубликована: Ноя. 22, 2024
Abstract
Artificial
intelligence
(AI)-based
multi-modal
fusion
algorithms
are
pivotal
in
emulating
clinical
practice
by
integrating
data
from
diverse
sources.
However,
most
of
the
existing
models
focus
on
designing
new
modal
methods,
ignoring
critical
role
feature
representation.
Enhancing
representativeness
can
address
noise
caused
heterogeneity
at
source,
enabling
high
performance
even
with
small
datasets
and
simple
architectures.
Here,
we
introduce
DeepOmix-FLEX
(Fusion
Learning
Enhanced
representation
for
X-modal
or
FLEX
short),
a
model
that
integrates
data,
proteomic
metabolomic
pathology
images
across
different
scales
modalities,
advanced
learning
contains
Feature
Encoding
Trainer
structure
train
encoding,
thus
achieving
inter-feature
inter-modal.
achieves
mean
AUC
0.887
prediction
chronic
kidney
disease
progression
an
internal
dataset,
exceeding
0.727
using
conventional
variables.
Following
external
validation
interpretability
analyses,
our
demonstrated
favorable
generalizability
validity,
as
well
ability
to
exploit
markers.
In
summary,
highlights
potential
AI
integrate
optimize
allocation
healthcare
resources
through
accurate
prediction.
Frontiers in Endocrinology,
Год журнала:
2025,
Номер
16
Опубликована: Апрель 16, 2025
Introduction
Our
study
aims
to
analyze
the
relationship
between
different
stage
of
Cardiovascular-Kidney-Metabolic
(CKM)
Syndrome
in
Chronic
Kidney
Disease
(CKD)
patients
and
risk
progression
all-caused
mortality
or
end-stage
renal
disease
(ESRD).
Methods
results
A
retrospective
cohort
was
performed
by
collecting
baseline
data
CKD
patients.
All
participants
were
followed
throughout
course
study.
Cox
proportional
hazards
analysis
Fine-Gray
subdistribution
model
prognostic
value
CKM
stages
on
adverse
clinical
outcomes
(all-caused
ESRD)
these
1,358
finally
completed
follow-up.
Among
them,
1,233
alive,
125
had
died;
163
progressed
ESRD.
Baseline
3
(OR=3.906,
95%
CI=0.988-16.320,
p=0.048)
4
(OR=5.728,
CI=1.329-24.698,
p=0.019)
remain
independent
factors
for
all-cause
patients,
while
2b
(OR=2.739,
CI=1.157-6.486,
p=0.022)
identified
as
having
an
factor
ESRD
adjusting
confounding
factors.
Conclusion
research
demonstrated
that
a
high-risk
can
predict
including
Journal of Clinical Medicine,
Год журнала:
2025,
Номер
14(8), С. 2833 - 2833
Опубликована: Апрель 19, 2025
Cardiovascular,
renal,
and
metabolic
diseases
are
pathophysiologically
interdependent,
posing
a
significant
global
health
challenge
being
associated
with
substantial
increase
in
morbidity
mortality.
In
2023,
the
American
Heart
Association
(AHA)
defined
this
complex
network
of
interconnected
conditions
as
cardiovascular–kidney–metabolic
(CKM)
syndrome.
This
syndrome
is
based
on
common
pathophysiological
mechanisms,
including
chronic
inflammation,
oxidative
stress,
hyperglycemia
insulin
resistance,
activation
renin–angiotensin–aldosterone
system
(RAAS),
neurohormonal
dysfunction,
which
trigger
vicious
cycle
where
impairment
one
organ
contributes
to
progressive
deterioration
others.
An
integrated
approach
these
conditions,
rather
than
treating
them
separate
entities,
supports
holistic
management
strategy
that
helps
reduce
burden
public
improve
patients’
quality
life.
Existing
focuses
lifestyle
modification,
glycemic
lipid
control,
use
nephroprotective
cardioprotective
therapies.
narrative
review
aims
synthesize
contextualize
existing
information
interactions
between
systems
diagnostic
approaches,
well
provide
an
overview
available
therapeutic
options.
Frontiers in Medicine,
Год журнала:
2024,
Номер
11
Опубликована: Окт. 18, 2024
In
China,
diabetes
mellitus
(DM)
significantly
contributes
to
end-stage
renal
disease
(ESRD),
necessitating
treatments
like
hemodialysis.
This
study
investigates
hemodialysis
outcomes
in
diabetic
nephropathy
patients
Guizhou
Province,
aiming
enhance
care
for
this
high-risk
group.
Briefings in Bioinformatics,
Год журнала:
2024,
Номер
26(1)
Опубликована: Ноя. 22, 2024
Abstract
Artificial
intelligence
(AI)-based
multi-modal
fusion
algorithms
are
pivotal
in
emulating
clinical
practice
by
integrating
data
from
diverse
sources.
However,
most
of
the
existing
models
focus
on
designing
new
modal
methods,
ignoring
critical
role
feature
representation.
Enhancing
representativeness
can
address
noise
caused
heterogeneity
at
source,
enabling
high
performance
even
with
small
datasets
and
simple
architectures.
Here,
we
introduce
DeepOmix-FLEX
(Fusion
Learning
Enhanced
representation
for
X-modal
or
FLEX
short),
a
model
that
integrates
data,
proteomic
metabolomic
pathology
images
across
different
scales
modalities,
advanced
learning
contains
Feature
Encoding
Trainer
structure
train
encoding,
thus
achieving
inter-feature
inter-modal.
achieves
mean
AUC
0.887
prediction
chronic
kidney
disease
progression
an
internal
dataset,
exceeding
0.727
using
conventional
variables.
Following
external
validation
interpretability
analyses,
our
demonstrated
favorable
generalizability
validity,
as
well
ability
to
exploit
markers.
In
summary,
highlights
potential
AI
integrate
optimize
allocation
healthcare
resources
through
accurate
prediction.
