Transcript Presentation

Electrophysiology & fMRI
Neurons
Neural computation
Neural selectivity
Hierarchy of neural processing
Integration of information
Retinal ganglion cell
receptive fields
Integrate
V1 neuron receptive field
(Hubel & Wiesel)
Grandma cell vs. distributed
population
Sparse coding
Narrow selectivity
Grandma cell vs. distributed
population
Distributed coding
Broad selectivity
Cortical columns
Neighboring neurons often
share the same selectivity and
are strongly connected.
“units of computation/selectivity”
Why such redundancy?
True for early sensory cortices.
Neurovascular coupling
Iadecola et. al. 2007
Synaptic activity
Ions
moving
across the
membrane
“Input” versus “output” of a neuron
Energy consumption
Brain activity consumes 20%
of the body’s energy at rest.
Glucose + oxygen => ATP
80% of the brain’s ATP is
used to keep sodium
potassium gradient.
Energy consumption
The power required for neural “Signaling” is a sum of
both neural output (spikes) and input:
Output
Input
Lennie P. Curr Bio 2003
Neural activity costs
Input is more expensive than output.
Neurons spend more energy on “listening” than “talking”
Dogma: Neural output (firing rate) is the interesting part.
Hemodynamic changes
Neural input or output?
Combination of both?
Electrophysiology
Different techniques:
1. Intra-cellular recordings
2. Extra-cellular recordings
3. Fluorescence imaging
Different spatial resolutions:
1. Single neuron activity
2. Multi unit activity
3. Local field potential
Extra-cellular recordings
Separate the recorded signal
into different components.
High frequencies (>500 Hz):
Low frequencies (<100 Hz):
Simultaneous measurements
Measure simultaneous electrophysiology and fMRI and
compare.
Logothetis et. al. Nature 2001
Simultaneous measurements
Before separating electrophysiology into different components
Simultaneous measurements
LFP and BOLD
responses are
sustained while
MUA seems to
adapt very
quickly.
Anything strange?
BOLD – spiking dissociations
Several other studies have reported such dissociations…
Viswanathan et. al. Nat. Neurosci. 2007
Spatial Sampling
MUA is a local measure, summing neural spikes only of
neurons surrounding the immediate electrode tip.
LFP and BOLD are wider measures, summing
dendritic/synaptic activity several mm surrounding the
electrode.
Sampling bias
Multi unit activity is mainly generated by large layer 5
pyramidal cells.
These are the main “output”
neurons of the cortex.
LFP and BOLD sum across
all cell sizes in all layers.
Cortical structure
In cortex, 80% of a neurons output synapses are located
within 1 mm of its soma.
Strong recurrent innervation.
Only 6% of V1 synapses
(mostly layer 4) are from
thalamic neurons.
Input without output?
Increased LFP without spiking?
BOLD, LFP, and spikes
In general, in cortex, there is a strong relationship. Epilepsy
patients implanted with electrodes in auditory cortex:
Mukamel et. al. Science 2005
BOLD, LFP, and spikes
Neural activity correlated with fMRI:
BOLD, LFP, and spikes
Different LFP frequencies showed different relationship to
BOLD:
Optogenetics
Hyung Lee et. al. Nature 2010
Optogenetics
Inject virus into
motor cortex.
Axons of
infected cells
reach thalamus.
Stimulate in
motor cortex and
measure activity
in both locations.
During rest
What about spontaneous activity?
Shmuel et. al. HBM 2008
During rest
Significant correlations between
neural activity and BOLD during
rest…
Negative BOLD?
Shmuel et. al. Nat. Neurosci. 2006
Negative BOLD?
Negative BOLD?
Cerebellum
Subcortical brain areas
might demand more
caution.
Architecture is
different: no recurrent
innervation.
GABA agonist
halts PC spikes
BOLD continues
There is a difference
between input and
output.
BOLD coupled to input.
Caeser et. al. PNAS 2003
One more thought
The effects of neuro-modulators (caffeine, hormones,
noradrenalin, dopamine, serotonin, etc…) on
particulars of neural activity and neurovascular
coupling are unknown.
To the lab