IEEE Spectrum January, 2014 - 47

in mice and other animals
will be characterized, along
with their molecular and
genetic properties, their
locations, and ultimately
how they connect to other
cells, both individually and
in groups.
These network-spanning
wiring diagrams are
called connectomes,
and they're considered
crucial. Researchers now
understand that an act of
perception or cognition
doesn't rely on the neurons
in just one area of the
brain; instead, it involves
complicated neural circuits
that can weave through
multiple regions. Although
similar connection-mapping
projects already exist-
the first roundworm
connectome was published
in 1986, and the NIH's
Human Connectome
Project launched in 2009-
they don't approach the
comprehensiveness or
detail envisioned in the
NIH report.
To study cells and their
networks in new detail,
investigators will need
to rely on a whole new
toolbox, possibly including
microscopes that can
look at larger sections of a
brain than is now practical,
and "optical needles" that
can penetrate a brain's
deep tissues, rather than
just the outermost layers
where most imaging now
occurs. Going beyond
passive imaging, some new
methods enable researchers
to activate cells or whole
networks inside a living
animal, helping them

fUn faCts:
150 Trillion
Estimated
number of
connections that link the
neurons in the human brain

160 milliseconds
Human
response time to an
auditory stimulus

750 milliliTers
per minuTe
Typical blood
flow through the human
brain (15 percent of
cardiac output)

250 000
neurons
per minuTe
Approximate rate
of neuron creation
during early pregnancy

8-10 seconds
Time until
unconsciousness after loss of blood
supply to the brain

1.5 kilograms
(3 pounds)
Average weight
of an adult brain (about
2 percent of the body's weight)

understand cell roles and
network dynamics and
eventually linking them
to behavior. Optogenetic
techniques, for example,
deliver light-sensitive
molecules to target cells,
then stimulate them with
light pulsed through a fiberoptic implant.
The BRAIN Initiative
hasn't given up on its
original goal of recording
neuronal activity, which
will also require new tools.
One proposed method
would make use of flexible
sheets fashioned from
hundreds of thousands
of nanowire electrodes,
which would conform
to a brain's topography
while noninvasively
recording its activity.
Researchers continue
to debate how much to
record: Eavesdropping
on every single cell in

a nervous system isn't
feasible, but how many are
representative?
As of now, activity can
be directly recorded from
perhaps 100 neurons at
a time, their electrical
spikes measured by tiny
implanted electrodes.
At the other end of the
spectrum is functional
magnetic resonance
imaging, or fMRI, which
measures blood flow as a
proxy for activity in regions
containing millions of cells.
"There's a lot of information
in the middle scale that
we don't have," says Jack
Gallant, a neuroscientist at
the University of California,
Berkeley. "And it's not just
that you need to record
from every neuron. You
need to record for a long
period of time."
Measuring that activity
over time is the only way
SPECTRUM.IEEE.ORG

|

to study the extraordinary
dynamism of brains. There's
no such thing as a static map
of a brain. Its connections
are ever-fluctuating,
reconfiguring themselves
on the fly. We now know that
the connectome is capable
of extreme transformation,
as seen in the brains of
stroke victims who recover
functions typically linked to
now-damaged regions.
Dealing with the resulting
data will require still more
tools, of the mathematical
and conceptual varieties.
Researchers will need
algorithms and processing
techniques that allow
them to make sense of raw
data and to understand
how neurological activity
encodes information, just
as computer codes are so
much more than strings of
ones and zeros. "There are
certain kinds of data that we
get from the brain and don't
know how to analyze," says
Gallant. "Nonlinear systems
with feedback"-weather,
for example, or brains-
"are mathematically hard
systems to deal with. There
just aren't good tools for
dealing with them."
when the nIh IssUes Its
first BRAIN grants in 2014,
it will likely concentrate
on projects that match
the agency's strengths,
such as direct work on
animals. Once the NIH's
concentrations are clear,
the NSF will start doling out
its funds, perhaps focusing
on its own strengths, such
as biophysics and computer
continued on page 64

nORTh aMERICan

|

jan 2014

|

47
http://SPECTRUM.IEEE.ORG
Table of Contents for the Digital Edition of IEEE Spectrum January, 2014
