How many specimens make a sufficient training set for automated three-dimensional feature extraction?
Royal Society Open Science,
Год журнала:
2024,
Номер
11(6)
Опубликована: Июнь 1, 2024
Deep
learning
has
emerged
as
a
robust
tool
for
automating
feature
extraction
from
three-dimensional
images,
offering
an
efficient
alternative
to
labour-intensive
and
potentially
biased
manual
image
segmentation
methods.
However,
there
been
limited
exploration
into
the
optimal
training
set
sizes,
including
assessing
whether
artficial
expansion
by
data
augmentation
can
achieve
consistent
results
in
less
time
how
these
benefits
are
across
different
types
of
traits.
In
this
study,
we
manually
segmented
50
planktonic
foraminifera
specimens
genus
Menardella
determine
minimum
number
images
required
produce
accurate
volumetric
shape
internal
external
structures.
The
reveal
unsurprisingly
that
deep
models
improve
with
larger
eight
being
95%
accuracy.
Furthermore,
enhance
network
accuracy
up
8.0%.
Notably,
predicting
both
measurements
structure
poses
greater
challenge
compared
structure,
owing
low
contrast
differences
between
materials
increased
geometric
complexity.
These
provide
novel
insight
sizes
precise
diverse
traits
highlight
potential
enhancing
multivariate
images.
Язык: Английский
Morphological evolution in a time of phenomics
Paleobiology,
Год журнала:
2025,
Номер
unknown, С. 1 - 19
Опубликована: Март 11, 2025
Abstract
Organismal
morphology
was
at
the
core
of
study
biodiversity
for
millennia
before
formalization
concept
evolution.
In
early
to
mid-twentieth
century,
a
strong
theoretical
framework
developed
understanding
both
pattern
and
process
morphological
evolution,
50
years
since
founding
this
journal
capture
transformational
period
in
quantification
analytical
tools
estimating
how
diversity
changes
through
time.
We
are
now
another
inflection
point
with
availability
vast
amounts
high-resolution
data
sampling
extant
extinct
allowing
“omics”-scale
analysis.
Artificial
intelligence
is
accelerating
pace
phenomic
acquisition
even
further.
This
new
reality,
which
ability
obtain
quickly
outpacing
analyze
it
robust,
realistic
evolutionary
models,
brings
set
challenges.
Phylogenetic
comparative
methods
have
provided
insights
into
processes
generating
diversity,
but
reliance
on
molecular
resultant
exclusion
fossil
from
most
large
phylogenetic
trees
has
well-established
negative
impacts
analyses,
as
we
demonstrate
examples
standard
single-rate
mode-
rate-shift
recently
described
Ornstein-Uhlenbeck
climate
model.
Further
development
analysis
high-dimensional
needed,
existing
can
refine
our
expectations
evolution
generation
under
different
scenarios,
analyses
placental
skull
Cenozoic.
Fully
transitioning
omics
era
will
involve
automate
extraction
meaningful,
comparable
morphometric
images,
integrate
downstream
generate
robust
models
that
accurately
reflect
complexity
well-suited
data.
Combined,
these
advancements
solidify
emerging
field
phenomics
appropriately
center
around
deep-time
Язык: Английский
Assessing the application of landmark-free morphometrics to macroevolutionary analyses
BMC Ecology and Evolution,
Год журнала:
2025,
Номер
25(1)
Опубликована: Апрель 27, 2025
Abstract
The
study
of
phenotypic
evolution
has
been
transformed
in
recent
decades
by
methods
allowing
precise
quantification
anatomical
shape,
particular
3D
geometric
morphometrics.
While
this
effectiveness
morphometrics
demonstrated
thousands
studies,
it
generally
requires
manual
or
semi-automated
landmarking,
which
is
time-consuming,
susceptible
to
operator
bias,
and
limits
comparisons
across
morphologically
disparate
taxa.
Emerging
automated
methods,
particularly
landmark-free
techniques,
offer
potential
solutions,
but
these
approaches
have
thus
far
primarily
applied
closely
related
forms.
In
study,
we
explore
the
utility
automated,
for
macroevolutionary
analyses.
We
compare
an
application
Large
Deformation
Diffeomorphic
Metric
Mapping
(LDDMM)
known
as
Deterministic
Atlas
Analysis
(DAA)
with
a
high-density
morphometric
approach,
using
dataset
322
mammals
spanning
180
families.
Initially,
challenges
arose
from
mixed
modalities
(computed
tomography
(CT)
surface
scans),
addressed
standardising
data
Poisson
reconstruction
that
creates
watertight,
closed
surfaces
all
specimens.
After
standardisation,
observed
significant
improvement
correspondence
between
patterns
shape
variation
measured
landmarking
DAA,
although
differences
emerged,
especially
Primates
Cetacea.
further
evaluated
downstream
effects
on
analyses,
finding
both
produced
comparable
varying
estimates
phylogenetic
signal,
morphological
disparity
evolutionary
rates.
Our
findings
highlight
like
DAA
large
scale
studies
taxa,
owing
their
enhanced
efficiency.
However,
they
also
reveal
several
should
be
before
can
widely
adopted.
context,
outline
issues,
propose
solutions
based
existing
literature,
identify
avenues
research.
argue
incorporating
improvements,
analyses
could
expanded,
thereby
enhancing
scope
enabling
analysis
larger
more
diverse
datasets.
Язык: Английский
Ten recommendations for scanning foraminifera by X-ray computed tomography
Journal of Micropalaeontology,
Год журнала:
2025,
Номер
44(1), С. 107 - 117
Опубликована: Апрель 29, 2025
Abstract.
Marine
sediment
cores
uniquely
provide
a
temporally
high-resolution
and
well-preserved
archive
of
foraminifera
fossils,
which
are
essential
for
understanding
environmental,
ecological,
evolutionary
dynamics
over
geological
timescales.
Foraminifera
preserve
their
entire
ontogeny
in
fossilized
shells,
much
this
life
history
remains
hidden
from
view
under
light
microscope.
X-ray
microfocus
computed
tomography
(μCT)
imaging
individual
reveals
internal
chambers
pores
that
traditionally
view.
Their
volume,
shape,
growth
form
foundations
oceanographic
environmental
research.
Here,
we
present
set
10
recommendations
the
preparation
scanning
using
glue-,
gel-,
solvent-free
methods.
We
focus
on
primary
parameters
μCT
researcher
can
optimize
according
to
throughput,
signal-to-noise
ratio,
cost
requirements
generate
three-dimensional
(3D;
volumetric)
datasets.
showcase
effect
these
image
quality
through
repeated
scans
single
planktonic
foraminifer
varied
beam
power
energy,
detector
binning,
number
projections,
exposure
times.
In
our
case
study,
highest
resulted
widest
contrast
between
subject
interest
background,
allowing
easiest
threshold-based
segmentation
object
aiding
computers
automated
feature
extraction.
The
values
exhibit
significant
variability
across
individuals,
based
specific
needs
equipment
used,
unique
attributes
samples
consideration.
Our
motivation
with
paper
is
share
experience
offer
foundation
similar
studies.
Язык: Английский
Morphological simulation tests the limits on phenotype discovery in 3D image analysis
bioRxiv (Cold Spring Harbor Laboratory),
Год журнала:
2024,
Номер
unknown
Опубликована: Июль 2, 2024
In
the
past
few
decades,
advances
in
3D
imaging
have
created
new
opportunities
for
reverse
genetic
screens.
Rapidly
growing
datasets
of
images
knockouts
require
high-throughput,
automated
computational
approaches
identifying
and
characterizing
phenotypes.
However,
exploratory,
discovery-oriented
image
analysis
pipelines
used
to
discover
these
phenotypes
can
be
difficult
validate
because,
by
their
nature,
expected
outcome
is
not
known
Язык: Английский
Introducing SPROUT (Semi-automated Parcellation of Region Outputs Using Thresholding): an adaptable computer vision tool to generate 3D segmentations
bioRxiv (Cold Spring Harbor Laboratory),
Год журнала:
2024,
Номер
unknown
Опубликована: Ноя. 23, 2024
Abstract
The
increased
availability
of
3D
image
data
requires
improving
the
efficiency
digital
segmentation,
currently
relying
on
manual
labelling,
especially
when
separating
structures
into
multiple
components.
Automated
and
semi-automated
methods
to
streamline
segmentation
have
been
developed,
such
as
deep
learning
smart
interpolation,
but
require
pre-labelled
data,
specialized
hardware
software.
Deep
models
in
particular
often
creation
extensive
training
particularly
for
complex
multi-class
segmentations.
Here,
we
introduce
SPROUT,
a
novel,
computer
vision
method
providing
time-efficient
user-friendly
pipeline
segmenting
parcellating
data.
SPROUT
generates
seeds
(representing
parts
an
object)
based
specified
density
thresholds
erosion
connected
components,
achieve
element
separation.
Seeds
are
grown
obtain
fully-parcellated
We
compare
SPROUT’s
performance
that
interpolation
apply
it
diverse
datasets
demonstrate
utility
versatility
this
open-source
method.
Язык: Английский
Calcification and ecological depth preferences of the planktonic foraminifer Trilobatus trilobus in the central Atlantic
Royal Society Open Science,
Год журнала:
2024,
Номер
11(12)
Опубликована: Дек. 1, 2024
Understanding
the
controls
behind
calcification
and
distribution
of
planktonic
foraminifera
in
modern
ocean
is
important
when
these
organisms
are
used
for
palaeoceanographic
reconstructions.
This
study
combines
previously
reported
shell
mass
data
with
new
geochemistry,
light
microscopy
X-ray
micro-computed
tomography
analyses
to
dissect
various
parameters
Trilobatus
trilobus
shells
from
surface
sediments,
investigating
factors
influencing
their
biometry.
The
goal
understand
which
aspects
marine
environment
critical
vertical
this
species.
found
produce
larger,
thinner
overall
lighter
equatorial
regions
than
subtropical
gyre
regions,
where
up
4%
smaller,
more
60%
thicker
approximately
45%
heavier.
skeletal
percentage
together
other
metrics
(shell
weight
thickness)
depend
primarily
on
ambient
seawater
salinity
rather
carbonate
chemistry.
In
line
degree
calcification,
basis
geochemically
reconstructed
apparent
depths,
group
shallower
water
column
at
Equator
gyres,
while
its
habitat
deepens
between
extra-equatorial
sites.
Furthermore,
it
demonstrated
that
(central)
Atlantic,
occupies
a
density
layer
slightly
below
maximum
isopycnal
presumably
by
adjusting
properties.
Язык: Английский