Nature Communications,
Journal Year:
2024,
Volume and Issue:
15(1)
Published: Oct. 31, 2024
In
a
perfect
world,
scientists
would
develop
analyses
that
are
guaranteed
to
reveal
the
ground
truth
of
research
question.
reality,
there
countless
viable
workflows
produce
distinct,
often
conflicting,
results.
Although
reproducibility
places
necessary
bound
on
validity
results,
it
is
not
sufficient
for
claiming
underlying
validity,
eventual
utility,
or
generalizability.
this
work
we
focus
how
embracing
variability
in
data
analysis
can
improve
generalizability
We
contextualize
design
decisions
brain
imaging
be
made
capture
variation,
highlight
examples,
and
discuss
may
quality
Brain
lacks
accessible
ground-truth
approaches,
leading
varied
results
across
field.
Embracing
analytical
allow
researchers
enhance
findings
accelerate
progress.
bioRxiv (Cold Spring Harbor Laboratory),
Journal Year:
2024,
Volume and Issue:
unknown
Published: April 19, 2024
1.The
use
of
machine
learning
to
classify
diagnostic
cases
versus
controls
defined
based
on
ontologies
such
as
the
ICD-10
from
neuroimaging
features
is
now
commonplace
across
a
wide
range
fields.
However,
transdiagnostic
comparisons
classifications
are
lacking.
Such
important
establish
specificity
classification
models,
set
benchmarks,
and
assess
value
ontologies.
Imaging Neuroscience,
Journal Year:
2024,
Volume and Issue:
2, P. 1 - 26
Published: Jan. 1, 2024
Abstract
Empirical
studies
reporting
low
test–retest
reliability
of
individual
blood
oxygen-level
dependent
(BOLD)
signal
estimates
in
functional
magnetic
resonance
imaging
(fMRI)
data
have
resurrected
interest
among
cognitive
neuroscientists
methods
that
may
improve
fMRI.
Over
the
last
decade,
several
reported
modeling
decisions,
such
as
smoothing,
motion
correction,
and
contrast
selection,
BOLD
estimates.
However,
it
remains
an
empirical
question
whether
certain
analytic
decisions
consistently
individual-
group-level
fMRI
task
across
multiple
large,
independent
samples.
This
study
used
three
samples
(Ns:
60,
81,
119)
collected
same
(Monetary
Incentive
Delay
task)
two
runs
sessions
to
evaluate
effects
on
(intraclass
correlation
coefficient
[ICC(3,1)])
group
(Jaccard/Spearman
rho)
activity
data.
The
this
vary
four
categories:
smoothing
kernel
(five
options),
correction
(four
parameterizing
(three
contrasts
totaling
240
different
pipeline
permutations.
Across
all
pipelines,
median
ICC
are
low,
with
a
maximum
estimate
.43
–
.55
3
greatest
impact
similarity
Implicit
Baseline
contrast,
Cue
Model
parameterization,
larger
kernel.
Using
condition
meaningfully
increased
compared
using
Neutral
cue.
effect
was
largest
for
parameterization;
however,
improvements
came
at
cost
interpretability.
illustrates
MID
variable
small
samples,
higher
not
always
interpretability
estimated
signal.
Neuropsychopharmacology,
Journal Year:
2024,
Volume and Issue:
50(1), P. 37 - 51
Published: Aug. 8, 2024
Abstract
Research
into
the
brain
basis
of
psychopathology
is
challenging
due
to
heterogeneity
psychiatric
disorders,
extensive
comorbidities,
underdiagnosis
or
overdiagnosis,
multifaceted
interactions
with
genetics
and
life
experiences,
highly
multivariate
nature
neural
correlates.
Therefore,
increasingly
larger
datasets
that
measure
more
variables
in
cohorts
are
needed
gain
insights.
In
this
review,
we
present
current
“best
practice”
approaches
for
using
existing
databases,
collecting
sharing
new
repositories
big
data
analyses,
future
directions
neuroimaging
psychiatry
an
emphasis
on
contributing
collaborative
efforts
challenges
multi-study
analysis.
Nature Communications,
Journal Year:
2024,
Volume and Issue:
15(1)
Published: Oct. 31, 2024
In
a
perfect
world,
scientists
would
develop
analyses
that
are
guaranteed
to
reveal
the
ground
truth
of
research
question.
reality,
there
countless
viable
workflows
produce
distinct,
often
conflicting,
results.
Although
reproducibility
places
necessary
bound
on
validity
results,
it
is
not
sufficient
for
claiming
underlying
validity,
eventual
utility,
or
generalizability.
this
work
we
focus
how
embracing
variability
in
data
analysis
can
improve
generalizability
We
contextualize
design
decisions
brain
imaging
be
made
capture
variation,
highlight
examples,
and
discuss
may
quality
Brain
lacks
accessible
ground-truth
approaches,
leading
varied
results
across
field.
Embracing
analytical
allow
researchers
enhance
findings
accelerate
progress.
