“Mimicking cortical responses in the visual cortex” Presentation 27 May 2004 Florie Daniels Lotte Verbunt.

“Mimicking cortical responsesin the visual cortex”

Presentation

27 May 2004

Florie DanielsLotte Verbunt

Introduction

Visual system is most important and well-known

One cell interactions between cells

Voltage sensitive dyes (Grinvald & Fitzpatrick)

Orientation preference

Introduction

Mapping respons dependent of stimulus in colour

500 m

Introduction

Spinning pinwheels

Goal: Reproducing the previous image and this movie in Mathematica

Contents

Biological backgroundModelling of simple cells and hypercolumns

of the cortex in MathematicaClusters: orientation of the hypercolumns in

relation to each otherTest-imagesColourmappingMovies Conclusions and suggestions

The optical pathway

The primary visual cortex

Hypercolumns

Processing a single 'pixel' in the visual

field

Orientation sensitivity

Scaling (sizes of receptive fields)

Receptive fields (RF)

Part of the visual field in which a stimulus will

elicit a respons

Small RF high resolution

Large RF blurred picture

Size RF = scale ()

fovea

The receptive field sensitivity profiles of simple cells

First order Gaussian derivative:

Second order Gaussian derivative:

φ = 0 and φ = π/2

φ = 0 and φ = π/2

The kernels for the hypercolumns

1st order

:14

1st order

:41

2nd order

:14

2nd order

:41

Clusters

Test-image: ramp

1st order

:14

2nd order

:14

1st order

:41

2nd order

:41

Test-image: line

1st order

:14

2nd order

:14

1st order

:41

2nd order

:41

Test-image: circle

1st order

:14

2nd order

:14

1st order

:41

2nd order

:41

Test-image: mr64

1st order

:14

2nd order

:14

1st order

:41

2nd order

:41

Colourmapping

The colourmapping

depends on:

angle colour

saturation 1

greyvalue brightness

Kernels in colour

mr64 in colour

Rotating bar (wide) in colour

Movie of a rotating line

Experiment vs model

Which part of cortex????

Rotation clustering

Rotating bar (wide)

Rotating bar (narrow)

Translating bar (wide)

Translating bar (narrow)

Conclusions

Goal not completely achieved, because of complexity and lack of time

σ increasing from the inside to the outside wide lines

σ increasing from the outside to the inside narrow lines

Second order Gaussian kernels are better line detectors than first order Gaussian kernels

Suggestions for further research

Mathematica:

Colour movie

Clustering

Experimental research:

Vary the stimuli

Tip:

Beware of mistakes in the x and y direction caused by plotting with

different plotting commands

Questions???