Diabetes Metabolic Syndrome and Obesity,
Journal Year:
2024,
Volume and Issue:
Volume 17, P. 1987 - 1997
Published: May 1, 2024
Purpose:
Diabetic
nephropathy
(DN),
a
major
complication
of
diabetes
mellitus,
significantly
impacts
global
health.
Identifying
individuals
at
risk
developing
DN
is
crucial
for
early
intervention
and
improving
patient
outcomes.
This
study
aims
to
develop
validate
machine
learning-based
predictive
model
using
integrated
biomarkers.
Methods:
A
cross-sectional
analysis
was
conducted
on
baseline
dataset
involving
2184
participants
without
DN,
categorized
based
their
development
over
follow-up
period
36
months:
(n=1270)
Non-DN
(n=914).
Various
demographic
clinical
parameters
were
analyzed.
The
findings
validated
an
independent
comprising
468
participants,
with
273
195
remaining
as
the
period.
Machine
learning
algorithms,
alongside
traditional
descriptive
statistics
logistic
regression
used
statistical
analyses.
Results:
Elevated
levels
serum
creatinine,
urea,
reduced
eGFR,
increased
prevalence
retinopathy
peripheral
neuropathy,
prominently
observed
in
those
who
developed
DN.
Validation
further
confirmed
model's
robustness
consistency.
SVM
demonstrated
superior
performance
training
set
(AUC=0.79,
F1-score=0.74)
testing
(AUC=0.83,
F1-score=0.82),
outperforming
other
models.
Significant
predictors
included
presence
diabetic
retinopathy,
neuropathy.
Conclusion:
Integrating
algorithms
biomarker
data
offers
promising
avenue
identifying
type
2
patients
36-month
Keywords:
nephropathy,
prediction,
learning,
biomarkers,
stratification,
Journal of Diabetes Research,
Journal Year:
2024,
Volume and Issue:
2024, P. 1 - 13
Published: Jan. 20, 2024
The
aim
of
this
study
is
to
analyze
the
effect
serum
metabolites
on
diabetic
nephropathy
(DN)
and
predict
prevalence
DN
through
a
machine
learning
approach.
dataset
consists
548
patients
from
April
2018
2019
in
Second
Affiliated
Hospital
Dalian
Medical
University
(SAHDMU).
We
select
optimal
38
features
least
absolute
shrinkage
selection
operator
(LASSO)
regression
model
10-fold
cross-validation.
compare
four
algorithms,
including
extreme
gradient
boosting
(XGB),
random
forest,
decision
tree,
logistic
regression,
by
AUC-ROC
curves,
calibration
curves.
quantify
feature
importance
interaction
effects
predictive
Shapley
additive
explanation
(SHAP)
method.
XGB
has
best
performance
screen
for
with
highest
AUC
value
0.966.
also
gains
more
clinical
net
benefits
than
others,
fitting
degree
better.
In
addition,
there
are
significant
interactions
between
duration
diabetes.
develop
algorithm
DN.
C2,
C5DC,
Tyr,
Ser,
Met,
C24,
C4DC,
Cys
have
great
contribution
can
possibly
be
biomarkers
Journal of Diabetes Science and Technology,
Journal Year:
2022,
Volume and Issue:
17(1), P. 224 - 238
Published: Sept. 19, 2022
Artificial
intelligence
can
use
real-world
data
to
create
models
capable
of
making
predictions
and
medical
diagnosis
for
diabetes
its
complications.
The
aim
this
commentary
article
is
provide
a
general
perspective
present
recent
advances
on
how
artificial
be
applied
improve
the
prediction
six
significant
complications
including
(1)
gestational
diabetes,
(2)
hypoglycemia
in
hospital,
(3)
diabetic
retinopathy,
(4)
foot
ulcers,
(5)
peripheral
neuropathy,
(6)
nephropathy.
JMIR Medical Informatics,
Journal Year:
2023,
Volume and Issue:
11, P. e47833 - e47833
Published: Oct. 12, 2023
Machine
learning
(ML)
models
provide
more
choices
to
patients
with
diabetes
mellitus
(DM)
properly
manage
blood
glucose
(BG)
levels.
However,
because
of
numerous
types
ML
algorithms,
choosing
an
appropriate
model
is
vitally
important.In
a
systematic
review
and
network
meta-analysis,
this
study
aimed
comprehensively
assess
the
performance
in
predicting
BG
In
addition,
we
assessed
used
detect
predict
adverse
(hypoglycemia)
events
by
calculating
pooled
estimates
sensitivity
specificity.PubMed,
Embase,
Web
Science,
Institute
Electrical
Electronics
Engineers
Explore
databases
were
systematically
searched
for
studies
on
levels
or
detecting
using
models,
from
inception
November
2022.
Studies
that
different
DM
included.
no
derivation
metrics
excluded.
The
Quality
Assessment
Diagnostic
Accuracy
tool
was
applied
quality
included
studies.
Primary
outcomes
relative
ranking
prediction
horizons
(PHs)
specificity
events.In
total,
46
eligible
meta-analysis.
Regarding
levels,
means
absolute
root
mean
square
error
(RMSE)
PH
15,
30,
45,
60
minutes
18.88
(SD
19.71),
21.40
12.56),
21.27
5.17),
30.01
7.23)
mg/dL,
respectively.
neural
(NNM)
showed
highest
PHs.
Furthermore,
positive
likelihood
ratio
negative
8.3
(95%
CI
5.7-12.0)
0.31
0.22-0.44),
respectively,
hypoglycemia
2.4
1.6-3.7)
0.37
0.29-0.46),
hypoglycemia.Statistically
significant
high
heterogeneity
detected
all
subgroups,
sources
heterogeneity.
For
precise
RMSE
increases
rise
PH,
NNM
shows
among
models.
Meanwhile,
current
have
sufficient
ability
events,
while
their
needs
be
enhanced.PROSPERO
CRD42022375250;
https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=375250.
Frontiers in Endocrinology,
Journal Year:
2024,
Volume and Issue:
15
Published: Feb. 28, 2024
Background
Peripheral
vascular
disease
(PVD)
is
a
common
complication
in
patients
with
type
2
diabetes
mellitus
(T2DM).
Early
detection
or
prediction
the
risk
of
developing
PVD
important
for
clinical
decision-making.
Purpose
This
study
aims
to
establish
and
validate
models
perform
factor
analysis
T2DM
using
machine
learning
Shapley
Additive
Explanation(SHAP)
based
on
electronic
health
records.
Methods
We
retrospectively
analyzed
data
from
4,372
inpatients
hospital
between
January
1,
2021,
March
28,
2023.
The
comprised
demographic
characteristics,
discharge
diagnoses
biochemical
index
test
results.
After
preprocessing
feature
selection
Recursive
Feature
Elimination(RFE),
dataset
was
split
into
training
testing
sets
at
ratio
8:2,
Synthetic
Minority
Over-sampling
Technique(SMOTE)
employed
balance
set.
Six
learning(ML)
algorithms,
including
decision
tree
(DT),
logistic
regression
(LR),
random
forest
(RF),
support
vector
machine(SVM),extreme
gradient
boosting
(XGBoost)
Adaptive
Boosting(AdaBoost)
were
applied
construct
models.
A
grid
search
10-fold
cross-validation
conducted
optimize
hyperparameters.
Metrics
such
as
accuracy,
precision,
recall,
F1-score,
G-mean,
area
under
receiver
operating
characteristic
curve
(AUC)
assessed
models’
effectiveness.
SHAP
method
interpreted
best-performing
model.
Results
RFE
identified
optimal
12
predictors.
XGBoost
model
outperformed
other
five
ML
models,
an
AUC
0.945,
G-mean
0.843,
accuracy
0.890,
precision
0.930,
recall
0.927,
F1-score
0.928.
importance
results
indicated
that
Hemoglobin
(Hb),
age,
total
bile
acids
(TBA)
lipoprotein(a)(LP-a)
are
top
four
factors
T2DM.
Conclusion
approach
successfully
developed
good
performance.
associated
offered
physicians
intuitive
understanding
impact
key
features
Journal of Diabetes Science and Technology,
Journal Year:
2024,
Volume and Issue:
18(2), P. 273 - 286
Published: Jan. 8, 2024
Importance
and
Aims:
Diabetic
microvascular
complications
significantly
impact
morbidity
mortality.
This
review
focuses
on
machine
learning/artificial
intelligence
(ML/AI)
in
predicting
diabetic
retinopathy
(DR),
kidney
disease
(DKD),
neuropathy
(DN).
Methods:
A
comprehensive
PubMed
search
from
1990
to
2023
identified
studies
ML/AI
models
for
complications.
The
analyzed
study
design,
cohorts,
predictors,
ML
techniques,
prediction
horizon,
performance
metrics.
Results:
Among
the
74
studies,
256
featured
internally
validated
124
had
externally
models,
with
about
half
being
retrospective.
Since
2010,
there
has
been
a
rise
use
of
complications,
mainly
driven
by
DKD
research
across
27
countries.
more
modest
increase
DR
DN
was
observed,
publications
fewer
For
all
predictive
achieved
mean
(standard
deviation)
c-statistic
0.79
(0.09)
internal
validation
0.72
(0.12)
external
validation.
highest
discrimination,
c-statistics
0.81
0.74
(0.13)
validation,
respectively.
Few
DN.
outcome
definitions,
number
type
technique
influenced
model
performance.
Conclusions
Relevance:
There
is
growing
global
interest
using
Research
most
advanced
terms
publication
volume
overall
Both
require
research.
External
adherence
recommended
guidelines
are
crucial.
Diabetes Metabolic Syndrome and Obesity,
Journal Year:
2024,
Volume and Issue:
Volume 17, P. 943 - 957
Published: Feb. 1, 2024
This
research
aims
to
examine
and
scrutinize
gender
variations
in
the
incidence
of
diabetic
nephropathy
(DN)
trajectory
renal
function
type
2
diabetes
mellitus
(T2DM)
patients.
Digital Health,
Journal Year:
2024,
Volume and Issue:
10
Published: Jan. 1, 2024
Previous
research
suggests
that
mathematical
models
could
serve
as
valuable
tools
for
diagnosing
or
predicting
diseases
like
diabetic
kidney
disease,
which
often
necessitate
invasive
examinations
conclusive
diagnosis.
In
the
big-data
era,
there
are
several
modeling
methods,
but
generally,
two
types
recognized:
conventional
model
and
machine
learning
model.
Each
method
has
its
advantages
disadvantages,
a
thorough
comparison
of
is
lacking.
this
article,
we
describe
briefly
compare
model,
provide
prospects
in
field.