Parkinson's
Disease
ranks
as
the
second
most
common
chronic
neurodegenerative
condition
that
affects
CNS
by
killing
cells
containing
dopamine
and
its
receptors.
Dopamine
is
responsible
for
coordination
controls
muscle
activity,
hence,
individuals
inflicted
with
do
unintended
or
involuntary
movements
due
to
lack
of
coordination.
Non-motor
symptoms
Disease,
also
known
as,
dopamine-non-responsive
encompass
issues
such
sleeping
difficulties,
constipation,
drooling,
swallowing
speech
impairments.
Notably,
90%
diseased
patients
suffer
from
impairments,
making
it
a
viable
sign
look
at
while
diagnosis.
Analyzing
acoustic
measurements
can
aid
in
early
diagnosis
enhancing
efficacy
treatment.
This
study
focuses
on
predicting
based
vocal
analysis
via
Machine
Learning
approach.
prediction
done
taking
into
account
various
metrics
like
frequency,
amplitude,
pitch,
intensity
tonality
undergo
alterations
Disease.
Speech-based
data
31
subjects
out
which
23
are
8
healthy
taken
create
points
testing
validation.
A
comparative
evaluation
machine
learning
models
an
ML-based
methodology
diagnose
individual
accuracy
96.15%
proposed
solely
basis
their
voice
structure
tonality.
Cognitive Computation,
Год журнала:
2024,
Номер
16(3), С. 1198 - 1209
Опубликована: Фев. 2, 2024
Abstract
Parkinson’s
disease
(PD)
is
a
neurological
condition
that
millions
of
people
worldwide
suffer
from.
Early
symptoms
include
slight
sense
weakness
and
propensity
for
involuntary
tremulous
motion
in
body
limbs,
particularly
the
arms,
hands,
head.
PD
diagnosed
based
on
motor
symptoms.
Additionally,
scholars
have
proposed
various
remote
monitoring
tests
offer
benefits
such
as
early
diagnosis,
ease
application,
cost-effectiveness.
patients
often
exhibit
voice
disorders.
Speech
signals
can
be
used
diagnosis
disease.
This
study
an
artificial
intelligence–based
approach
using
speech
signals.
Scalogram
images,
generated
through
Continuous
Wavelet
Transform
signals,
were
employed
deep
learning
techniques
to
detect
PD.
The
scalograms
tested
with
techniques.
In
first
part
experiment,
AlexNet,
GoogleNet,
ResNet50,
majority
voting-based
hybrid
system
classifiers.
Secondly,
feature
fusion
method
DenseNet
NasNet
was
investigated.
Several
evaluation
metrics
assess
performance.
achieved
accuracy
0.95
F1
score
stratified
10-fold
cross-validation,
improving
by
38%
over
ablation
study.
key
contributions
this
investigation
scalogram
images
comprehensive
analysis
models
detection.
Biomimetics,
Год журнала:
2023,
Номер
8(4), С. 351 - 351
Опубликована: Авг. 7, 2023
Parkinson’s
disease
(PD)
affects
a
large
proportion
of
elderly
people.
Symptoms
include
tremors,
slow
movement,
rigid
muscles,
and
trouble
speaking.
With
the
aging
developed
world’s
population,
this
number
is
expected
to
rise.
The
early
detection
PD
avoiding
its
severe
consequences
require
precise
efficient
system.
Our
goal
create
an
accurate
AI
model
that
can
identify
using
human
voices.
We
transformer-based
method
for
detecting
by
retrieving
dysphonia
measures
from
subject’s
voice
recording.
It
uncommon
use
neural
network
(NN)-based
solution
tabular
vocal
characteristics,
but
it
has
several
advantages
over
tree-based
approach,
including
compatibility
with
continuous
learning
network’s
potential
be
linked
image/voice
encoder
more
multi
modal
solution,
shifting
SOTA
approach
(NN)
crucial
advancing
research
in
multimodal
solutions.
outperforms
state
art
(SOTA),
namely
Gradient-Boosted
Decision
Trees
(GBDTs),
at
least
1%
AUC,
precision
recall
scores
are
also
improved.
additionally
offered
XgBoost-based
feature-selection
fully
connected
NN
layer
technique
measures,
addition
network.
discussed
numerous
important
discoveries
relating
our
suggested
deep
(DL)
application
such
as
how
resilient
increased
depth
compared
simple
MLP
performance
proposed
conventional
machine
techniques
MLP,
SVM,
Random
Forest
(RF)
have
been
compared.
A
detailed
comparison
matrix
added
article,
along
solution’s
space
time
complexity.
Journal of Information Systems Engineering and Business Intelligence,
Год журнала:
2024,
Номер
10(1), С. 38 - 50
Опубликована: Фев. 28, 2024
Background:
Parkinson's
disease
(PD)
is
a
critical
neurodegenerative
disorder
affecting
the
central
nervous
system
and
often
causing
impaired
movement
cognitive
function
in
patients.
In
addition,
its
diagnosis
early
stages
requires
complex
time-consuming
process
because
all
existing
tests
such
as
electroencephalography
or
blood
examinations
lack
effectiveness
accuracy.
Several
studies
explored
PD
prediction
using
sound,
with
specific
focus
on
development
of
classification
models
to
enhance
The
majority
these
neglected
crucial
aspects
including
feature
extraction
proper
parameter
tuning,
leading
low
Objective:
This
study
aims
optimize
performance
voice-based
through
extraction,
goal
reducing
data
dimensions
improving
model
computational
efficiency.
Additionally,
appropriate
parameters
will
be
selected
for
enhancement
ability
identify
both
cases
healthy
individuals.
Methods:
proposed
new
applied
an
OpenML
dataset
comprising
voice
recordings
from
31
individuals,
namely
23
patients
8
participants.
experimental
included
initial
use
SVM
algorithm,
followed
by
implementing
PCA
machine
learning
Subsequently,
balancing
SMOTE
was
conducted,
GridSearchCV
used
best
combination
based
predicted
characteristics.
Result:
Evaluation
showed
impressive
accuracy
97.44%,
sensitivity
100%,
specificity
85.71%.
excellent
result
achieved
limited
10-fold
cross-validation
rendering
sensitive
training
data.
Conclusion:
successfully
enhanced
SVM+PCA+GridSearchCV+CV
method.
However,
future
investigations
should
consider
number
folds
small
dataset,
explore
alternative
methods,
expand
generalizability.
Keywords:
GridSearchCV,
Parkinson
Disaese,
SVM,
PCA,
SMOTE,
Voice/Speech
