Lecture Kerkoerle Flashcards
Detailed neural mechanisms of top-down visual processing in macaque monkeys
Cortical hierarchy
buttom-up stream in hirarchy
visual imput –> V1/V2 (spatial location) different neurons represent different features. –> V4 (larger receptive fields - more complex) differentiation in different elements (color, curved, rectilinear) –> TEO/TE –? vIPF
Receptive fields
V1/V2 neurons in certain location that represent different visual elements (one location, differing neurons representing differing features). Neighbouring neurons that represent neighbouring locations in the visual field
V4
larger receptive fields (more complex, i.e. circular or t-shaped) separation into several feature dimensions. I.e. Color, Curved, Rectilinerar
TEO/TE
Visual fields become even bigger (entire objects, such as banana, apple, etc). Still division into several features (color, curved, rectilinear)
vIPFC
completely abstract representations (i.e. circle, color, rectilinear,…)
Artificial neural network (ANN)
convolutional neural network (filter)
what is the directionality of the cortical hierarchy
it is actually 2-way, bottom-up is well understood, while top-down is not very well understood.
selection of visual information in depression/schizophrenia, etc. iss impaired
First need to understand healthy functioning, to then understand what the impairment is
How can top-down processing be investigated?
- Laminar recordings
- multi-photon imaging
- distribution of cell types
Why work with primates?
- primates are genetically vey similar to humans
- cell-types : genetic profile of important neurons (PVALB, SST, LAMPS, VIP)
- In macaque and Marmoset the expression level of these neurons shows high correlation with the human one
- translation is not possible if working with mouse or ferret
How is the organization of macaque different to mouse
Anatomical connectivity between different areas with association cortex. In visual area there is a clear hierarchy.
Large network dissociated from sensory input, you can have internal representations, like working memory that is not affected by sensory imput
In the mouse there is almost no hierarchy, everything is more connected. less high level cognitive functions
How is the columnal organization in V1 different between monkeys, mice and humans?
in V1 you have columnal organization, different orientations for different neurons which are clustered in macaque monkeys
but in rodents there is no such organization
Monkeys vs humans
human- 3x bigger
V1 is relatively smaller in humans compared to macaque (prefrontal cortex has increased, but V1 is relatively shrunken.
This is not the case if you “count the neurons” –> in humans there is no difference to other monkeys.
Human cortex has increased in volume, but the number of neurons has not increased.
BUT- Neurons have become bigger
Basically, there is a different in degree, not in kind
it correlate in a difference in behavior (working memory)
Human capacity to memorize (ca. 4 elements), in monkeys for 2 items it already drops about 80%, this mean a monkey usually can only memorize about 1 element
Laminar recording
Feedforward and feedback connections target separate cortical layers.
V1 : tracing individual axons
feedforward connections from LGN -> Layer 4A, L.4C, L.6
feedback connections: Axons coming from V1, arriving in L.1, L.2, and a little in L.3 and L.5 (avoids 4 and 6)
separate input coming in to the different layers (L4, L6 and L.3,L.5,L.1/2) in order to separate feedforward and feedback connections
Laminar recordings in monkey V1
- Flash evoked activity
- presenting high contrast checkboard grading
- record simultaneously spiking activity across different layers
MUA
- normalization, all peaks are now 1 to see differences in time : L6 and L4 is a little bit earlier
- Synaptic input spreads very quickly, which is why it is hard to identify it
CSD: 2nd
spiking activity (action potentials)
LFP (corresponds to synaptic activity ) x2. subtract the neighbours
–> currents flowing away or towards the channel
current sink (where current flows inside neurons) - L4C - Feedforward
LFP vs action potential
LFP: summed activity of a population of neurons (in Hz)
Action potentials: individual neural firing
Selective attention task
4 possible targets
curve tracing task
monkey fixates a red dot. Then lines appear connecting red dot to one of 4 targets
Monkey has to follow curve and make eye movement to the target
First you record at the target, then the distractor
–> need for larger receptive fields
target vs distractor
Spiking activity across layers
fixation (300 ms)
Stimulus (750 ms)
Saccade (monkey gets juice if correct)
if stimulus appears, there is a very strong response (peak) that can propagate to the next layer
Around 200 miliseconds: stabilizes in time.
Feedforward (no distinction between target and distractor)
only at about 200 milliseconds a distinction takes place
Feedback laminar profile
MUS and cortical depth and CSD shows feedback in the correct layers
Is V1 involved in working memory as well?
Stimulus 150 ms, but followed by delay so that monkey has to memorize.
Neurons in V1 still have modulation between target and distractor
There is working memory trace in V1 but lost if mask is presented
Working memory is only stored in prefrontal cortex
Multi-photon imaging & activity-dependent fluorophores advantages
- sub-cellular resolution (ca. 2um)
- complete neural population
- map topography
- follow neurons over sessions (see same neuron and detect learning effect)
- connections between neurons/areas
- cell-type specificity
- image neuromodulators (e.g. dopamine)
link anatomy with congnitive functions
Dissecting the neural microcircuit
A. Local somata and projections
Injection Site
Viral Expression
B. Cell-type specificity
C. Anterograde
D. Retrograde
You engineer a virus, it infects a cell and the cell starts producing any protein you want it to produce
local expression of virus
cell-type specific (inject virus anywhere and only specific neurons will be affected)
Viral vector technology
Multi-photon microscopy
Light scattering–> overcome bad resolution
one photons vs 2 photons vs 3 photons excitation : the more light you get, the more signal you get
only at focal plane you get excitation
multi-photoimaging in monkeys
visualizing neuronal connections using 2-photon imaging
Is there changing in V1
window implanted –> image axons and understand how they change while animal is learning the task.
Before learning and after learning axons changed while animal was learning
after a couple of weeks the dura grew back
if you have 3-photon imaging you can reach deeper
Three-microscope suitable for macaque monkeys
Imaging vasculature in monkey cortex
Imaging neurons in monkey cortex
measure calcium activity
2/ vs 3/photon imaging of the vasculature in the cortex through the dura
section the brain with light
Distribution of cell types
Comparing hierarchies between mice and macaque monkeys
- cortical gradient of interneuron distribution
Feedforward -> conects to perimidal cells and PV cells
Feedback-> CR projecting to (inhibits) CB projecting to (disinhibiting) pyramidal
in macaque there is V2/V1 feedforward and ACC feedback
Mouse only has feedfarward processes
Summary
- Selective attention involves feedback to V1
- Working memory involves feedback to V1, but working memory items are not stored in V1
- Three-photon microscopy allows imaging through natural dura in macaque monkeys
- Top-down circuits become dominant towards higher cortical areas, and this gradient is enhanced in macaque monkeys compared to mice
- This suggests that primates might have unique abilities to maintain internal goals and dynamically select information that is relevant for the task at hand
