29
Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp [email protected] HowYourBrainWorks.net

Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp [email protected] HowYourBrainWorks.net

Embed Size (px)

Citation preview

Page 1: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Principles of Sensory Neuroscience

Systems Biology Doctoral Training ProgramPhysiology course

Prof. Jan [email protected]

HowYourBrainWorks.net

Page 2: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Transduction

Fig 2 of http://www.masseyeandear.org/research/ent/ent-investigators/eatock/

Page 3: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Labelled Line Codes

Already before the specific tuning properties of sensory receptors could be demonstrated, Rene Descartes hypothesized that sensory afferents carry modality specific information to the brain.Microstimulation of single cutaneous afferents in human volunteers links specific nerve fibres to specific sensations (e.g. RA-II fibre activation causes the sensation of “flutter”)

Page 4: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Place Codes

• Ramon y Cajal speculated in the 1930s that the optic chiasm may have evolved to allow an uninterrupted topographic representation of the visual scene on the surface the optic tectum. (A, B)

• We now know that topographic maps in the tectum are discontinuous. (C).

• The notion that topographic maps contribute to neural representations remains widespread.

Page 5: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

The Receptive Field Concept

Fig 22.3 of Kandel et al “Principles of Neural Science”

Page 6: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

A Better Receptive Field Concept…

… describes the behaviour of a sensory neuron quantitatively in terms of a “transfer function” y=f(x) which maps a mathematical description of the stimulus x (location, intensity, frequency, colour, temperature, recent history …) onto a measure of the neuron’s “output” y (depolarization, firing probability, response latency).

Page 7: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Rate Codes

• Classic experiments performed by Adrian in the 1920s on frog muscle stretch receptors established that sensory afferents use changes in spike rate to signal the intensity of a stimulus.

• Adrian also found that many sensory neurons “adapt”, i.e. they do not maintain very high firing rates for long if stimuli are held constant.

0 1 2 3Weight (g)

0

50

100

150

Rat

e (H

z)

0 1 2 3Weight (g)

0

50

100

150

Rat

e (H

z)

Page 8: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Quantifying Rate Codes: The Post Stimulus Time Histogram

Source: http://www.frontiersin.org/Journal/10.3389/fnsyn.2010.00017/full

Page 9: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Eye and Retina

Page 10: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Centre –Surround Receptive Fields+

-

--

--

++

+

-

--

---

Photo-receptors

HorizontalCell

BipolarCell

RetinalGanglionCell

Page 11: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Difference of Gaussians Model of Retinal Ganglion Cells

• The centre-surround structure of Retinal Ganglion Cells turns them into “spatial frequency filters”. Larger RGC receptive fields are tuned to “coarsely grained” structure in the visual scene, while smaller RFs are tuned to fine grain structure.

Page 12: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Convolving a Penny with DoGs

• The picture of an American cent (left) seen through large (middle) or small (right) difference of Gaussian receptive fields.

Page 13: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Seeing Lines

Page 14: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

The Gabor Filter Model of V1 Simple Cells

Retina->LGN->V1 simple cell: linearRetina->LGN->V1 simple cell: linear+ -+++

--- --

+--

+-

+-

-

+

+--

-

LGN Cortex

Page 15: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Linear Filters in Visual Cortex?

• The “F0/F1” ratio is often used to distinguish simple (approximately linear) V1 neurons from complex (nonlinear) ones.

• Responses are recorded to sinusoidal contrast gratings. If the cell is linear, the output should contain only the input frequency F0.

• Fourier analysis is performed on the post stimulus time histogram to measure the amplitude ratio of the fundamental (1st harmonic, F1) to the “zero frequency” (i.e. sustained, “DC” response) F0.

• Some complex cells have “on” and “off” responses which manifest themselves as F2=2·F1 components - a “quadratic” (... +c·sin(x)2 +...) non-linearity.

Movshon, Tolhurst and Thompson 1978Movshon, Tolhurst and Thompson 1978

Page 16: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Pennies as seen by V1 simple cells

• American cent coin (original to the left) convolved with “Gabor” simple cell receptive field models shown above.

Page 17: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

The Ear

Organ of Corti

Cochlea “unrolled” and sectioned

Page 18: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

“Gammatone Filter Bank”

Page 19: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Auditory Nerve Fibers behave like Rectified Gammatone Filters

Auditory Neuroscience Fig 2.12Based on data collected by Goblick and Pfeiffer (JASA 1969) 

Page 20: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

The Auditory Pathway

M GB

IC

NLL

SOC

CN

Cor

tex

C och lea

M GB

IC

NLL

SOC

CN

Cortex

C och lea

Bra

inst

emM

idbr

ain

CN, cochlear nuclei; SOC, superior olivary complex; NLL, nuclei of the lateral lemniscus; IC, inferior colliculus; MGB, medial geniculate body.

Page 21: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

From work by Shihab Shamma and colleagues

Linear Neural Filters In Auditory Cortex?

Page 22: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Measuring Frequency-Time (Spectro-Temporal) Receptive Fields with Reverse Correlation

time

Fre

q. c

hann

el

Page 23: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Binaural Frequency-Time Receptive Field

Page 24: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Linear Prediction

of Responses

-5 0 5 10

1

4

16

dB

1

4

16

Fre

qu

enc

y [k

Hz]

r(t) = i1(t-1) w1(1) + i1(t-2) w1(2)+ ...+ i2(t-1) w2(1) + i2(t-2) w2(2)+ ...+ i3(t-1) w3(1) + i2(t-2) w3(2)+ ...

Latency

FTRF “w matrix”

Input“i vector”

01002000

0.5

1re

spo

nse

ms

Page 25: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

01002000

0.5

1

resp

on

se

ms 0 100 2000

200

rate

(H

z)

ms

-5 0 5 10

1

4

16

dB

1

4

16

Left and Right Ear Frequency-Time Response

FieldsVirtual Acoustic Space Stimuli

Fre

qu

en

cy

[kH

z]

a

c

d

e

f

b

C81

-180 -120 -60 0 60 120 180Azim [deg]

-60

0

60

Ele

v [d

eg]

Ele

v [

deg

]

Predicting Space from Spectrum

Schnupp et al Nature 2001Schnupp et al Nature 2001

Page 26: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Are Neurons “Noisy” Rate Coders? Or Precision Spike Timers?

Mainen & Sejnowski, Science 1995

Page 27: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

What about Spike Latency

Codes?• Many nervous in the central

auditory system seem to fire only short bursts of action potentials at the onset of a stimulus.

• For such neurons, the response latency may vary as a function of certain stimulus parameters (e.g. intensity, sound source position … ) and could therefore encode that parameter.

• Nelken et al, “Encoding stimulus information by spike numbers and mean response time in primary auditory cortex” J Comput Neurosci (2005)

So

und

so

urce

azi

mu

th ()

Page 28: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

The discharges of cochlear nerve fibres to low frequency sounds are not random; they occur at particular times (phase locking).The spike time intervals therefore encode temporal features of the stimulus (sound periodicity).

The discharges of cochlear nerve fibres to low frequency sounds are not random; they occur at particular times (phase locking).The spike time intervals therefore encode temporal features of the stimulus (sound periodicity).

Evans (1975)Evans (1975)

How about Spike Interval Codes?

Page 29: Principles of Sensory Neuroscience Systems Biology Doctoral Training Program Physiology course Prof. Jan Schnupp jan.schnupp@dpag.ox.ac.uk HowYourBrainWorks.net

Who reads the neural code, and

how do they do it?