Receptive fields and neural analysis Flashcards
What are receptive fields?
The area of the sensory surfaces which, if stimulated, will affect a neuron’s membrane potential, threshold or firing rate.
E.g receptive fields overlap and vary with size, in the eye receptive fields are larger in the periphery where visual acuity is less of a priority relative to the fovea in the centre of the eyes where receptive fields are very small.
Size of receptive field is inversely proportional to acuity and the subsequent cortical area associated with that receptive field (RF).
Ganglion cells in the retina have concentric receptive fields with on-centres and off-surrounds or off-centres and on-surrounds.
Depending on where light falls in the spatial structure of these concentric-opponent receptive fields determines the cells’ response - fire/not fire.
What is the spatial structure of a receptive field?
Excitatory on-center and inhibitory off-surround.
Light shone onto the on-centre of receptive field results in the cells’ firing rate increasing about the spontaneous rate
Light shone onto the off-surround results in the cells’ firing rate decreasing below the spontaneous rate.
If light is shone on both simultaneously, then the excitation and inhibition cancel out and there is no change in firing rate.
Polarity is reversed for off-centre, on-surround RFs
What is the circuitry of receptive fields?
Photoreceptors (rods and cones) line the back of the retina
When light interacted with their photopigments, the photoreceptors become hyperpolarised.
Photoreceptor in centre of RF → bipolar cells → ganglion cells
Photoreceptor in surround of RF → horizontal cells → inhibit bipolar cells
The sum of excitation and inhibition is conveyed to the ganglion cell which varies its firing rate appropriately.
All of the signals in the circuit are encoded by graded potentials up to the ganglion cells.
Ganglion cells transmit action potentials to the optic nerve which convey the combined signals from all the ganglion cells to the brain.
What is the linearity of receptive fields?
Linear RF - cells response is the sum of the excitation and inhibition in the entire receptive field
Non-linear RF - the response is not the sum of excitation and inhibition and has no distinct on or off regions.
What is the different classification of cells?
Properties and nomenclature vary depending on whether referring to cats or monkeys
Cat’s have no colour vision and have X and Y cells
X cells have small RFs, are linear, sustained response, broadband
Y cells have large RFs, are non-linear, transient responses, broadband.
Monkeys have colour vision and P and M cells
P cells have small linear RFs, with sustained responses, wavelength
M cells have large non-linear RFs and are more transient and broadband
X and Y cells resemble M cells
As cats are colourblind they have nothing that resembles P cells.
What were Hubel and Wiesel’s experiments with cats in the optic nerve?
Visual fields can be plotted by inserting a microelectrode through a cat’s skull into ganglion cells and measuring the varying firing rate.
A dot of light is moved on a screen
When it falls in the middle of the receptive field there is a burst of activity indicating an excitatory on response.
When the light is off and there is a burst of activity this indicates an inhibitory off response.
Can conclude there are receptive fields with on-centres and off-surrounds and vice versa.
On-centre RF, light falls on cone cell in centre of RF → excitatory response to bipolar cell → graded potential to ganglion cell.
Off-surround RF, light falls on cone cell in off-surround → excitatory response to horizontal cells → inhibit bipolar cells → no activation of ganglion cells. This is called lateral inhibition.
When there is light in both the centre and surround the response is much weaker due to the lateral inhibition effect.
Ganglion cells can respond much better to local differences of intensity than overall illumination of the eye.
In large receptive fields, lots of photoreceptors converge onto one bipolar cell. Many of these bipolar cells converge onto one ganglion cell. Result is greater sensitivity to light at the cost of acuity.
On-centre, off-surround RFs respond best to light dot against a dark background
Off-centre, on-surround RFs respond best to dark dot against a light background.
What were Hubel and Wiesel’s experiments with cats in the visual cortex?
Optic nerve →LGN → visual cortex
Measured cortical cells’ change in firing rate in response to light to map receptive fields.
Found cells with increasingly complex receptive fields.
Simple cells: will always respond to a long bar of light with on-zones and off-zones
Orientation detectors as changing the orientation results in light falling simultaneously on the on and off zones so the neuron produces no response.
Complex cells: larger RFs, more complex connections, often spontaneously active.
Direction sensitive as they respond to moving light in the correct orientation rather than flashing light or an illuminated bar.
Hyper complex cells: size detectors as they respond to bars of a set width and length, that are moving and are the correct orientation.
From this Hubel and Wiesel hypothesised that the visual cortex is organised into columns of interconnected cells and that all the neurons in a column respond to the same orientation. Simple cells feed into complex cells which feed into hypercomplex cells.
If a microelectrode is driven down perpendicular to the surface of a column, every neuron is found to respond to the same orientation where if it penetrated diagonally there are sudden changes in preferred orientation between columns.
What is the visual pathway?
Ganglion cells in the reina project onto relay neurons in the lateral geniculate nucleus (LGN)
Relay cells as their RFs are similar to ganglion cells opponent-concentric.
Some have sustained temporal responses and linear spatial responses: X-cells
Some have transient temporal responses and non-linear spatial responses: Y-cells
Relay neurons project ono simple cells in the primary visual cortex
Orientation selective with linear responses.
Simple cells project onto complex cells
Orientation selective, direction selective and have non-linear responses
Complex cells project to hypercomplex cells
Orientation selective, direction selective and non-linear spatial responses that are end stopped so selective for length.
How did Barlow support Rfs and templates?
Originated with Barlow’s (1953) suggestion that the off-centre, on-surround RF in frog’s retina would make ideal fly detectors as they respond best to a dark dot against a white background.
Thus the concentric-surround RF provides a template against which a pattern of light can be matched.
If it is matched then the target will be detected.
How did Hubel and Wiesel argue for RFs as feature detectors?
Extended Barlow’s theory to account for complex and hypercomplex cells.
RFs at each level of the system can be constructed by combining the outputs of RF from the previous level to make successively more complex RFs.
Simple cells: converging outputs from a row of relay cells with concentric-opponent RFs onto a cell in the cortex that responds maximally to a bar of light that straddles the row of relay cells. This makes it an orientation detector.
Complex cells: converging outputs of a row of simple cells onto one cell in the cortex that responds to a bar of light moving across all the simple cells. Direction and orientation selective.
Hypercomplex cells: convergent outputs of complex cells.
This can be extrapolated so increasingly complex cells with larger receptive fields with more specific response properties at each successive level.
What did Gross’s experiments with monkeys find?
Removing parts of the inferior temporal (IT) cortex in monkeys affects their ability to discriminate between different objects.
Recorded single neurons in the monkey’s IT cortex over a period of 4 days
Neurons did not fire in response to simple stimuli such as lines, squares, circles light or dark.
Neurons do respond to more complex stimuli. Neurons spiked in response to an accidental shadow of a hand across the screen.
Found the neuron responded maximally to a handlike shape with fingers pointing up.
Some neurons responded best to faces.
The concept that neurons respond to real life objects was revolutionary.
Kanwisher et al. (1997) found the fusiform gyrus in humans to respond optimally to faces hence the ‘fusiform face area’.
Neurons respond to complex real world stimulim - could support Hubel and Wiesel’s hypothesis that neurons have increasingly specialised RFs.
What are the criticisms of the template model?
circuitry timing gnostic units combinatorial explosion principle of univariance
What is the problem with circuitry in the template model?
No one has been able to trace the circuitry
There is far more overlap and convergence between the LGN and striate cortex than Hubel and Wiesel’s model implies.
Fails to account for feedback circuitry (not just feed-forward connections).
explain how timing is a problem for the template model
Latencies of response of simple, complex and hypercomplex cells do not conform to the sequence implied by the model
These empirical observations count against Hubel and Wiesel’s model specifically but do not mean that template matching is not a good model, rather that it is not a good model of template matching.
explain the gnostic units criticism
Gnostic units - Konorski (1967)
The logical end point of successively combine features to be increasingly elaborate is a gnostic unit.
A gnostic unit is so specific it responds to one highly complex pattern exclusively.
Lettvin (1969) ridiculous these gnostic units by referring to them as grandmother cells claiming that they are so specific they could respond only to the image of a person’s grandmother.
Coding visual information like this would render percepts and memories fragile as they would be lost with the cell death of one neuron.
This would contradict the graceful degradation neural networks exhibit in practise.
