Nature Methods,
Journal Year:
2024,
Volume and Issue:
21(5), P. 908 - 913
Published: March 21, 2024
Mapping
neuronal
networks
from
three-dimensional
electron
microscopy
(3D-EM)
data
still
poses
substantial
reconstruction
challenges,
in
particular
for
thin
axons.
Currently
available
automated
image
segmentation
methods
require
manual
proofreading
many
types
of
connectomic
analysis.
Here
we
introduce
RoboEM,
an
artificial
intelligence-based
self-steering
3D
'flight'
system
trained
to
navigate
along
neurites
using
only
3D-EM
as
input.
Applied
mouse
and
human
cortex,
RoboEM
substantially
improves
state-of-the-art
segmentations
can
replace
more
complex
analysis
problems,
yielding
computational
annotation
cost
cortical
connectomes
about
400-fold
lower
than
the
error
correction.
In
most
animals,
a
relatively
small
number
of
descending
neurons
(DNs)
connect
higher
brain
centers
in
the
animal’s
head
to
circuits
and
motor
(MNs)
nerve
cord
body
that
effect
movement
limbs.
To
understand
how
signals
generate
behavior,
it
is
critical
these
pathways
are
organized
onto
MNs.
fly,
Drosophila
melanogaster
,
MNs
controlling
muscles
leg,
wing,
other
systems
reside
ventral
(VNC),
analogous
mammalian
spinal
cord.
companion
papers,
we
introduced
densely-reconstructed
connectome
Male
Adult
Nerve
Cord
(MANC,
Takemura
et
al.,
2023),
including
cell
type
developmental
lineage
annotation
(Marin
which
provides
complete
VNC
connectivity
at
synaptic
resolution.
Here,
present
first
look
organization
networks
connecting
DNs
based
on
this
new
information.
We
proofread
curated
all
ensure
accuracy
reliability,
then
systematically
matched
DN
axon
terminals
MN
dendrites
with
light
microscopy
data
link
their
morphology
inputs
or
muscle
targets.
report
both
broad
organizational
patterns
entire
network
fine-scale
analysis
selected
interest.
discover
direct
DN-MN
connections
infrequent
identify
communities
intrinsic
linked
control
different
systems,
putative
for
walking,
dorsal
flight
steering
power
generation,
intermediate
lower
tectulum
coordinated
action
wings
legs.
Our
generates
hypotheses
future
functional
experiments
and,
together
MANC
connectome,
empowers
others
investigate
richer
mechanistic
detail.
Nature,
Journal Year:
2024,
Volume and Issue:
630(8017), P. 686 - 694
Published: June 5, 2024
Abstract
To
convert
intentions
into
actions,
movement
instructions
must
pass
from
the
brain
to
downstream
motor
circuits
through
descending
neurons
(DNs).
These
include
small
sets
of
command-like
that
are
sufficient
drive
behaviours
1
—the
circuit
mechanisms
for
which
remain
unclear.
Here
we
show
DNs
in
Drosophila
directly
recruit
networks
additional
orchestrate
require
active
control
numerous
body
parts.
Specifically,
found
previously
thought
alone
2–4
fact
co-activate
larger
populations
DNs.
Connectome
analyses
and
experimental
manipulations
revealed
this
functional
recruitment
can
be
explained
by
direct
excitatory
connections
between
interconnected
brain.
Descending
population
is
necessary
behavioural
control:
with
many
partners
network
co-activation
complete
only
simple
stereotyped
movements
their
absence.
DN
reside
within
behaviour-specific
clusters
inhibit
one
another.
results
support
a
mechanism
generated
increasingly
large
compose
combining
multiple
subroutines.
Nature Methods,
Journal Year:
2024,
Volume and Issue:
21(5), P. 908 - 913
Published: March 21, 2024
Mapping
neuronal
networks
from
three-dimensional
electron
microscopy
(3D-EM)
data
still
poses
substantial
reconstruction
challenges,
in
particular
for
thin
axons.
Currently
available
automated
image
segmentation
methods
require
manual
proofreading
many
types
of
connectomic
analysis.
Here
we
introduce
RoboEM,
an
artificial
intelligence-based
self-steering
3D
'flight'
system
trained
to
navigate
along
neurites
using
only
3D-EM
as
input.
Applied
mouse
and
human
cortex,
RoboEM
substantially
improves
state-of-the-art
segmentations
can
replace
more
complex
analysis
problems,
yielding
computational
annotation
cost
cortical
connectomes
about
400-fold
lower
than
the
error
correction.
