Transcript Some Principles of Stimulus Evoked Cortical Dynamics of Visual Areas

SOME PRINCIPLES OF STIMULUS EVOKED
CORTICAL DYNAMICS OF VISUAL AREAS
Per.E.Roland
Bashir Ahmed
Michel Harvey
Akitoshi Hanazawa
Calle Undeman
David Eriksson
Sarah Wehner
Sonata Valentiniene
Brain Research, Dept. Neuroscience, Karolinska Institute , Stockholm,
Sweden
Neuron computations start by afferent inputs to the synapses (pre- and postsynaptic), propagate into the
dendrites, which perform nonlinear operations, and end by producing electrical spike activity, action
potential (AP), or no action potentials .
The result of the computation is a spike train. Neurons communicate by APs and
transmitter diffusion. No single neuron can drive the brain.
Roland 2002
How do single neurons work together and at which scale ?
CORTICAL DYNAMICS
Definition: in vivo spatial and temporal organization of
computations
and communications
by cortical neurons in real time
Complex dynamic systems are characterized by their
Architecture (invariant for shorter time periods)
And
Their dynamics
Transients induce dynamics which is different from dynamic states
One cannot predict the dynamics form the architecture
Ferret brain (mustela putorius) working at the mesoscopic scale in vivo
We stain the cortex with a Voltage sensitive dye
The voltage sensitive dye binds to the
membranes of all neurons. When the
membrane depolarizes, the dye changes
conformation < 1s and emit
fluorescence at a higher wavelength
Antic et al 1999
1. A STATIONARY
OBJECT
Stimulus a 133 ms luminance contrast square
25 ms
50 ms
83 ms
133 ms
250 ms
No stim
Single trial: luminance contrast square exposed for 133 ms, start 0
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
A Small square lasting 83 ms
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
Time derivative of population membrane potentials = C inward current
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
Laminar recording area 17/18 to stationary square in center of field of view
The feedback wave
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
Neurons from area 21,19 and 18 fire to the feedback wave
p < 0.0001
Roland et al. 2006
Single stationary square
The excitatory connexions in the cerebral cortex (Roland 2008)
The spike train elicited by a luminance contrast defined object interacts with the
ongoing activity in area 17 and evokes
1.
Thalamo-cortical feed-forward firing IV spreading to III and II and inducing
a (relative) depolarization in area 17. The onset of firing in the layers goes
in the order IV, III, V, II and VI.
2.
Lateral spreading of the (relative) depolarization and firing of neurons
representing the object background, continuing until feedback (4)
3.
Feed-forward (relative) depolarization of areas 19 and 21
4.
With a further delay a Feedback wave of (relative) depolarization of most
of areas 19,18 and 17 interacting first with the neurons at the 17 object representation
to increase and then decrease the membrane potential here and apparently segment
the object from background
5
a spreading decrease of excitation from the area 17 object representation
6
And presumably a second broad feed-forward excitation of area 18,19, 21 and higher
The visual system computes scenes rather than objects
2. OBJECT MOTION
UP FROM PERIPHERY
DOWN FROM PERIPHERY
Object
Background
x,y
Retina stationary
All that is mapped on the cortex is mapped with a Delay 40 ms
So how can animals & humans ever catch or avoid an object?
t3-t4
x,y ds/dt
t1-t2
A MOVING OBJECT WILL BE MAPPED IN MANY VISUAL AREAS
Ferret visual cortex
2 x 1O bar moving upwards
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
Harvey et al subm
UP FROM PERIPH
DOWN FROM PERIPH
1. STATIONARY
Membrane potentials form layers I-III; Firing from layer IV
DOWN FROM PERIPH
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
25º/sec 824ms
Membrane potentials from layers I-III. Firing from layers V-VI
DOWN FROM PERIPH
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
25º/sec 824ms
Moving objects on the retina are mapped, with a delay of ~50 ms,
moving in retinotopic organized visual areas.
Area 17/18 send feed-forward the object motion to areas 19/21 (layer IV).
In the examples of linear motion, area 19/21 compute a prediction
of the future trajectory of the object after ~ 130 ms.
This prediction is sent as feedback to area 17 (layers V VI) instructing
area 17 neurons to compute similar prediction and predepolarizing the
future cortical path.
The prediction maps the future position 250 ms ahead of the
object’s position in cortex.
This gives the animal (human) sufficient time to saccade or prepare
and execute limb movement.
Meanwhile, the object mappings move over the cortex in phase, due to
the predepolarization in area 17
3. APPARENT MOTION
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
Apparent
motion
Ahmed et al. 2008
Apparent motion, population membrane potential
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
Ahmed et al. 2008
Ahmed et al. 2008
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
The square is first mapped as stationary until 116 ms
Split motion
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
Ahmed et al. 2008
d(V(t)rel,AM-V(t)rel;sum)/dt or the difference in dynamics between AM signal
and the sum of signals to stationary single squares at identical positions and times
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
Ahmed et al. 2008
Ahmed et al. 2008
Signals that humans perceive as moving objects. When the identical
Square stimuli are shown to the ferret, The square stimuli are
initially mapped in area 17 as stationary, but
Time-locked to the offset of the first square
1. The mapping of the square in area 19/21 moves towards the
second square
2. A feedback signal from area 19/21 instructs area 17 to
depolarize the path in the direction of apparent motion and
3. The mapping of the square in area 17 moves towards the site
of the next square
The mapping of the square in 19/21 was computed as moving,
but computed as stationary in area 17. This discrepancy elicit a
feedback from the higher order area forcing are 17 to compute object
motion
General conclusions (so far)
At the mesoscopic scale, the cerebral cortex is well behaved
In real time studies
Communications are reflected in changes of the membrane
potentials of the target populations of neurons
Examples of communications: feed-forward, feedback with
different messages, lateral spreading depolarizations.
Higher order areas may enslave lower order areas though
feedback.
The lateral spreading depolarizations and the feedbacks
engage very large neuron populations in all visual areas so
far measured.
For stationary objects the feed-forward -feedback
computations are finished < 120-130 ms. For moving objects
the computations and communications goes on.
Temporal derivative of population membrane potentials, d(∆V(t))/dt, of all animals
aligned to cytoarchitectural borders: area 19/21 teaching area 17 the prediction
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
A: single square at 3 positions in 3 different trials
B: apparent motion, square successively at the 3 positions initially mapped as stationary
The offset of
short duration stimuli
elicit a decrease in
the inward current
that postpones the
OFF response
r(t) firing rate
Consequently
The time interval
Between the ON
and OFF firing peak
Is prolonged for
Stimuli < 100 ms